Library Review Digital libraries and the semantic web: context, applications and research

lr cover (i).qxd

21/04/2008

13:36

Page 1

ISSN 0024-2535

Volume 57 Number 3 2008

Library Review Digital libraries and the semantic web: context, applications and research Guest Editor: George Macgregor

www.emeraldinsight.com

Library Review

ISSN 0024-2535 Volume 57 Number 3 2008

Digital libraries and the semantic web: context, applications and research Guest Editor George Macgregor

Access this journal online _________________________

171

Editorial advisory board___________________________

172

GUEST EDITORIAL Introduction to a special issue on digital libraries and the semantic web: context, applications and research George Macgregor _____________________________________________

173

ANTAEUS The practitioner librarian and the semantic web Nicholas Joint _________________________________________________

178

Integrated access to cultural heritage resources through representation and alignment of controlled vocabularies Antoine Isaac, Stefan Schlobach, Henk Matthezing and Claus Zinn ______

187

The AGROVOC Concept Server: rationale, goals and usage Margherita Sini, Boris Lauser, Gauri Salokhe, Johannes Keizer and Stephen Katz __________________________________________________

Access this journal electronically The current and past volumes of this journal are available at:

www.emeraldinsight.com/0024-2535.htm You can also search more than 175 additional Emerald journals in Emerald Management Xtra (www.emeraldinsight.com) See page following contents for full details of what your access includes.

200

CONTENTS

CONTENTS continued

Reducing semantic complexity in distributed digital libraries: treatment of term vagueness and document re-ranking Philipp Mayr, Peter Mutschke and Vivien Petras _____________________

213

Faceted infrastructure for semantic digital libraries A.R.D. Prasad and Devika P. Madalli ______________________________

225

Semantic heterogeneity: comparing new semantic web approaches with those of digital libraries Ju¨rgen Krause_________________________________________________

235

Book reviews _____________________________________

249

www.emeraldinsight.com/lr.htm As a subscriber to this journal, you can benefit from instant, electronic access to this title via Emerald Management Xtra. Your access includes a variety of features that increase the value of your journal subscription.

How to access this journal electronically To benefit from electronic access to this journal, please contact [email protected] A set of login details will then be provided to you. Should you wish to access via IP, please provide these details in your e-mail. Once registration is completed, your institution will have instant access to all articles through the journal’s Table of Contents page at www.emeraldinsight.com/0024-2535.htm More information about the journal is also available at www.emeraldinsight.com/ lr.htm Our liberal institution-wide licence allows everyone within your institution to access your journal electronically, making your subscription more cost-effective. Our web site has been designed to provide you with a comprehensive, simple system that needs only minimum administration. Access is available via IP authentication or username and password. Emerald online training services Visit www.emeraldinsight.com/training and take an Emerald online tour to help you get the most from your subscription.

Key features of Emerald electronic journals Automatic permission to make up to 25 copies of individual articles This facility can be used for training purposes, course notes, seminars etc. This only applies to articles of which Emerald owns copyright. For further details visit www.emeraldinsight.com/ copyright Online publishing and archiving As well as current volumes of the journal, you can also gain access to past volumes on the internet via Emerald Management Xtra. You can browse or search these databases for relevant articles. Key readings This feature provides abstracts of related articles chosen by the journal editor, selected to provide readers with current awareness of interesting articles from other publications in the field. Reference linking Direct links from the journal article references to abstracts of the most influential articles cited. Where possible, this link is to the full text of the article. E-mail an article Allows users to e-mail links to relevant and interesting articles to another computer for later use, reference or printing purposes. Structured abstracts Emerald structured abstracts provide consistent, clear and informative summaries of the content of the articles, allowing faster evaluation of papers.

Additional complimentary services available Your access includes a variety of features that add to the functionality and value of your journal subscription: Xtra resources and collections When you register your journal subscription online you will gain access to additional resources for Authors and Librarians, offering key information and support to subscribers. In addition, our dedicated Research, Teaching and Learning Zones provide specialist ‘‘How to guides’’, case studies and interviews and you can also access Emerald Collections, including book reviews, management interviews and key readings. E-mail alert services These services allow you to be kept up to date with the latest additions to the journal via e-mail, as soon as new material enters the database. Further information about the services available can be found at www.emeraldinsight.com/alerts Emerald Research Connections An online meeting place for the world-wide research community, offering an opportunity for researchers to present their own work and find others to participate in future projects, or simply share ideas. Register yourself or search our database of researchers at www.emeraldinsight.com/connections

Choice of access Electronic access to this journal is available via a number of channels. Our web site www.emeraldinsight.com is the recommended means of electronic access, as it provides fully searchable and value added access to the complete content of the journal. However, you can also access and search the article content of this journal through the following journal delivery services: EBSCOHost Electronic Journals Service ejournals.ebsco.com Informatics J-Gate www.j-gate.informindia.co.in Ingenta www.ingenta.com Minerva Electronic Online Services www.minerva.at OCLC FirstSearch www.oclc.org/firstsearch SilverLinker www.ovid.com SwetsWise www.swetswise.com

Emerald Customer Support For customer support and technical help contact: E-mail [email protected] Web www.emeraldinsight.com/customercharter Tel +44 (0) 1274 785278 Fax +44 (0) 1274 785201

LR 57,3

EDITORIAL ADVISORY BOARD

AFRICA Dr Priti Jain Librarian and Guest Lecturer, University of Botswana

172

AUSTRALASIA Dr G.E. Gorman, FCLIP FRSA Professor of Library and Information Management, School of Information Management, Victoria University of Wellington, New Zealand Jake Wallis Lecturer, Charles Sturt University, Australia EAST ASIA Dr Kuan Yew Wong Department of Manufacturing and Industrial Engineering, Faculty of Mechanical Engineering, Universiti Teknologi Malaysia EUROPE Robin Adams Librarian, Trinity College Library Dublin Dr Mary A. Burke Head of School, UCD School of Information and Library Studies, University College Dublin Dr Jose-Antonio Gomez-Hernandez Assistant Professor, Faculty of Information and Communication Studies, University of Murcia, Spain Jette Hyldegaard The Royal School of Librarianship, Copenhagen, Denmark Sirje Virkus Head, Department of Information Studies, Tallinn University, Estonia AMERICAS Dr S. Nazim Ali Director of the Finance and Banking Information Project, Centre for Middle Eastern Studies, Harvard University, Cambridge, MA, USA Dr Judith Licea de Arenas Professor of Library Science, Faculty of Philosophy and Letters, University of Mexico, Mexico City Catherine Cardwell Associate Professor, Chair, Library Teaching and Learning, University Libraries, Bowling Green State University, Bowling Green, OH, USA Dr Christie M. Koontz College of Information, The Florida State University, USA

Library Review Vol. 57 No. 3, 2008 p. 172 # Emerald Group Publishing Limited 0024-2535

Dr Jesu´s Lau Director, USBI VER Library, Universidad Veracruzana, Veracruz, Me´xico Professor Ali Shiri School of Library and Information Studies, University of Alberta, Edmonton, Canada UNITED KINGDOM Susan Ashworth Library, University of Glasgow, UK Briony Birdi Lecturer, MA Programme Coordinator, Department of Information Studies, University of Sheffield, UK Dr M.E. Burke Information Systems Institute, University of Salford, UK Gobinda Chowdhury Senior Lecturer, Department of Computer and Information Science, University of Strathclyde, UK Professor Forbes Gibb Lecturer, Department of Computer and Information Sciences, University of Strathclyde, Glasgow UK Alan Gilchrist Consortium and TFPL Ltd, UK Dick Hartley Head, Department of Information and Communications, Manchester Metropolitan University, UK J.D. Hendry County Heritage Services Officer, Cumbria County Council, Carlisle, UK Stuart James Former Librarian, University of Paisley, Scotland, UK Jenny Rowley Lecturer, School for Business and Regional Development, University of Wales, Bangor, UK Dr Jane Secker Learning Technology Librarian, London School of Economics, UK Professor Patricia Layzell Ward Gwynedd, Wales, UK Sheila Webber Lecturer, Department of Information Studies, Sheffield University, UK

The current issue and full text archive of this journal is available at www.emeraldinsight.com/0024-2535.htm

GUEST EDITORIAL

Introduction to a special issue on digital libraries and the semantic web: context, applications and research

Guest editorial

173

George Macgregor Information Strategy Group, Liverpool Business School, Liverpool John Moores University, Liverpool, UK Abstract Purpose – The purpose of this article is to introduce the papers in the special issue which explores some of the potential, opportunities and challenges to be found in greater library and information science alignment with semantic web developments. Design/methodology/approach – The article is a general review of the papers in the issue. Findings – For many digital libraries or cultural institutions, the semantic web offers an opportunity to better expose valuable digital resources pertaining to research, culture or history, using common standards and technologies in a collaborative and ‘‘joined up’’ way. The papers in this issue ‘‘paint a rainbow’’, exploring the issues through elements of case studies, reviews research and conceptual expositions and viewpoints. Originality/value – The article emphasises how the practical implications of semantic web research or developments for digital libraries and repositories is important for LIS professionals. Keywords Digital libraries, Worldwide web, Information management Paper type Viewpoint

Digital library: Digital libraries are organisations that provide the resources, including the specialised staff, to select, structure, offer intellectual access to, interpret, distribute, preserve the integrity of, and ensure the persistence over time of collections of digital works so that they are readily and economically available for use by a defined community or set of communities (Digital Library Federation, 1998).

The Semantic Web: The Semantic Web is not a separate Web but an extension of the current one, in which information is given well-defined meaning, better enabling computers and people to work in cooperation. The first steps in weaving the Semantic Web into the structure of the existing Web are already under way. In the near future, these developments will usher in significant new functionality as machines become much better able to process and ‘‘understand’’ the data that they merely display at present (Berners-Lee et al., 2001).

Digital libraries are now a ‘‘mature information service application’’ (Bearman, 2007), the parameters of which have been delineated by continuous definition and conceptualisation, almost since the advent of the web itself. By contrast, the Semantic The author wishes to extend his thanks to David McMenemy (Library Review editor) for his editorial guidance throughout the stewardship of this special issue, and to all those who kindly participated as peer reviewers. Their involvement was greatly appreciated.

Library Review Vol. 57 No. 3, 2008 pp. 173-177 # Emerald Group Publishing Limited 0024-2535 DOI 10.1108/00242530810865457

LR 57,3

174

Web remains a project in its infancy; an attempt to augment resources on the web with metadata about their semantics, not just their syntax. Making the web as readable to machines as it currently is to humans is the Holy Grail. The prospect of a machinereadable web creates the potential for a plethora of intelligent, information rich applications, the scope of which precludes a dozen special issues (see Legg, 2007 for a brief review). There clearly exists numerous opportunities for digital libraries to use what the Semantic Web can offer. Perhaps the most seductive of these relates to improved user resource discovery across distributed heterogeneous collections (via search disambiguation or inference) and improved data interoperability, among many others. This special issue presents a collection of papers aimed at exploring some of the potential, opportunities and challenges to be found in greater LIS alignment with Semantic Web developments. For many digital libraries or cultural institutions, the Semantic Web offers an opportunity to better expose valuable digital resources pertaining to research, culture or history, using common standards and technologies in a collaborative and ‘‘joined up’’ way. Semantic Web technologies are capable of enhancing digital libraries or repositories by facilitating improved navigation and retrieval within heterogeneous document environments, user profiling, personalisation and contextualisation, improved user interfaces and human-computer interaction. Such technologies also have the potential to solve or aid the management of problems relevant to digital libraries and the LIS community generally, such as semantic interoperability, advanced metadata and information integration, the management of large corpora of heterogeneous digital resources, and so forth. Given the potential benefits to be gleaned and the clear synergy between LIS and the Semantic Web, much of the digital library community has been slow to assimilate Semantic Web developments, or has chosen to ignore them altogether. Whether this has been a consequence of a perception that it is the preserve of the W3C and irrelevant to the community, or whether both communities are talking at cross-purposes remains a moot point. Fortunately, it is possible to write that this position has been altering over recent years. Greenberg and Me´ndez (2007a) suggest the change was precipitated by the emerging notion of ‘‘Library 2.0’’, first proposed by Miller (2005) and further developed in 2006 (Miller, 2006). Miller’s ‘‘call to arms’’ may use Web 2.0 as the hook, but its detail has awoken a wider recognition within LIS that it has to be more proactive in delivering valuable content to users and must – more generally – participate in the evolution of the web itself. This special issue was announced in early 2007. Since then Greenberg and Me´ndez (Greenberg and Me´ndez, 2007b) have edited a special issue of Cataloging and Classification Quarterly on a similar theme, published simultaneously as an indispensable monograph. This special issue of Library Review attempts to continue the collection of papers presented by Greenberg and Me´ndez and build upon similarly themed conferences (Prasad and Madalli, 2007), the latter of which originally inspired this special issue of Library Review. Since this issue is using previous literature as a springboard, some basic Semantic Web concepts may elude exposition. Those readers requiring further detail are encouraged to consult Greenberg and Me´ndez (2007b), part I of which includes papers on the Semantic Web building blocks, or Legg (2007).

Digital libraries and the Semantic Web: context, applications and research The papers contained in this special issue comprise a rainbow painted in the splendour of the Emerald paper categories[1]; encompassing elements of case studies, general reviews, research and conceptual expositions, and viewpoints. This issue begins with a Library Review staple: the ANTAEUS column. Nick Joint provides musings on the special issue theme and considers the impact of Semantic Web developments on digital libraries and the practitioner librarian. Joint outlines some of the broad issues associated with the Semantic Web, together with a simple explanation of some basic Semantic Web principles. He concludes that the Semantic Web is of fundamental importance to LIS practitioners (particularly for digital repository developments and the increased exposure of open access materials) and argues that if the true Semantic Web dream is to be realised, the LIS community will have to play an active role. However – in true ANTAEUS style – Joint also provides a cautionary note and reminds us that the futuristic nature of Semantic Web/digital library interactions requires practitioner-oriented research to ensure the implementation of meaningful and practical applications. The LIS community has much in common with developments in the Semantic Web. This is perhaps most obvious in the Semantic Web use of ontology; an area that resonates squarely with the study of other Knowledge Organisation Systems (KOS), such as classification schemes, taxonomies, thesauri or subject heading lists. Most of the papers in the special issue explore – or at least touch on – the use of such techniques as a means of addressing issues of semantic heterogeneity. Emanating from the STITCH project[2] and highly active in the development of the Semantic Web itself, Antoine Isaac, Henk Matthezing, Stefan Schlobach and Claus Zinn demonstrate how Semantic Web techniques can ameliorate semantic interoperability issues within the cultural heritage domain, providing users with integrated and seamless access to heterogeneous collections. Isaac et al. deploy the Simple Knowledge Organisation System (SKOS)[3] and automatic ‘‘vocabulary alignment’’ methods to facilitate semantic user searching and navigation across collections. Whilst Isaac et al. concede that these techniques require further refinement and experimentation, they conclude that it is the future for digital libraries and cultural heritage collections; successfully shifting us from ‘‘separate islands of collections and vocabularies to better connected networks of cultural heritage knowledge’’. With a mission to combat hunger through improved knowledge exchange, the Knowledge Exchange and Capacity Building Division (KCE)[4] of the Food and Agriculture Organization of the United Nations (FAO)[5] continue to lead developments in digital library technologies. Margherita Sini, Boris Lauser, Gauri Salokhe, Johannes Keizer and Stephen Katz discuss the development of the AGROVOC concept server (CS). The OWL[6] compliant CS is designed as a ‘‘collaborative reference platform’’, providing a collection of commonly used agricultural concepts, containing a plethora of terms, definitions and relationships between terms in multiple languages. Sini et al. also detail the CS Workbench, a CS module enabling the distributed and collaborative management of the CS, in turn facilitating the re-use and extension of agricultural knowledge for increased interoperability and improved user services. Philipp Mayr, Peter Mutschke and Vivien Petras use the German science portal ‘‘vascoda’’[7] as a test-bed for implementing techniques designed to ameliorate user term vagueness and to improve result rankings for users. They demonstrate a search term recommender system employing terminology mappings and query expansion techniques. Approaches derived from scientometrics and network analysis are also

Guest editorial

175

LR 57,3

176

deployed to better rank user result sets. Noting various synergies with Friend-of-aFriend (FOAF)[8], Mayr et al. rank results according to the core journals of specific domains of knowledge (Bradfordizing) and by the centrality of authors in coauthorship networks. It is the conclusion of Mayr et al. that greater intersection between the digital library and the Semantic Web communities will produce improved analytical tools and interfaces for the presentation of information better adapted to users’ needs; something that Mayr et al. intend to pursue in further work. As we have noted, the LIS use and deployment of KOS resonates with the Semantic Web use of onotology. This, in turn, is an area that resonates with the work of S.R. Ranganathan, a founding father of modern library science and key founder of the Documentation Research and Training Centre (DRTC), Bangalore[9]. Continuing Ranganathan’s legacy and emanating from DRTC itself, Prasad and Devika Madalli use their paper to propose a faceted infrastructure model for semantic digital libraries. This conceptual model – to be implemented and evaluated as part of a wider research study – employs a ‘‘centre-out approach’’; the centre being semantic representations (based on faceted ontologies) of information. Prasad and Madalli also provide us with a useful reappraisal of why digital libraries have to be cognisant of Semantic Web developments and seize this new potential for enhanced user resource discovery. As Isaac et al. remind us, digital libraries offer access to large amounts of heterogeneous resources. Semantic Web approaches to the management of these resources (e.g. through the use of ontology to ensure the consistent labelling and description of resources and the use of common technical standards) can ameliorate or resolve the interoperability issues often encountered by digital libraries. Resolving heterogeneity to ensure interoperability or to improve information retrieval is the focus Ju¨rgen Krause, who picks up themes raised by Mayr et al. Krause closes the special issue with a detailed theoretical exposition, exploring the development of Semantic Web approaches in tandem with those of digital libraries. In particular, Krause focuses on the ‘‘Shell model’’, an approach used in the German science portal ‘‘vascoda’’, and conceptually analyses the two approaches which – although employing different techniques – seek the same goal (i.e. to resolve semantic heterogeneity). He places both under intense theoretical scrutiny, noting areas for optimism, but also those areas which require caution. Krause also notes the mismatch between the Semantic Web rationale for using controlled vocabularies instead of ontologies (by using alternatives such as SKOS, for example) and that of the LIS community, and argues that the both communities have to revert to ‘‘weaker semantic foundations’’ via SKOS in order to achieve the ultimate aim of the Semantic Web: interoperability. It was the intention of the guest editor that this issue should emphasise the practical implications of Semantic Web research or developments for digital libraries and repositories; for research to be given an applied LIS focus in order to increase relevance to the usual Library Review readership. Considering all the papers together, it is hoped that the reader develops a fuller appreciation of the special issue theme. And it is hoped that both the digital library and Semantic Web community benefit from this activity. At the very least it is hoped that it stimulates further comment on areas of useful overlap between communities and how the aims of each can be furthered by increased collaboration.

Notes 1. Guidelines for writing structured abstracts: www.emeraldinsight.com/info/authors/ writing_for_emerald/abstracts.jsp (accessed 01 December 2007). 2. STITCH Project: http://stitch.cs.vu.nl/ (accessed 01 December 2007). 3. Simple Knowledge Organisation Systems (SKOS): www.w3.org/2004/02/skos/ (accessed 01 December 2007). 4. Knowledge Exchanges and Capacity Building Division (KCE): www.fao.org/gi/gil/ index_en.asp (accessed 01 December 2007). 5. Food and Agriculture Organization of the United Nations (FAO): www.fao.org/ (accessed 01 December 2007). 6. Web Ontology Language (OWL): www.w3.org/2004/OWL/ (accessed 01 December 2007). 7. vascoda: http://www.vascoda.de/ (accessed 01 December 2007). 8. Friend-of-a-Friend (FOAF) Vocabulary Specification: http://xmlns.com/foaf/spec/ (accessed 01 December 2007). 9. Documentation Research and Training Centre (DRTC): http://drtc.isibang.ac.in/DRTC/ (accessed 01 December 2007). References Bearman, D. (2007), ‘‘Digital libraries’’, in Cronin, B. (Ed.), Annual Review of Information Science and Technology, Vol. 41, Information Today Inc., New Jersey, NJ, pp. 223-72. Berners-Lee, T., Hendler, J. and Lassila, O. (2001), ‘‘The Semantic Web’’, Scientific American Magazine, May, available at: www.sciam.com/article.cfm?articleID=00048144-10D2-1C7084A9809EC588EF21 (accessed 1 December 2007). Digital Library Federation (1998), ‘‘A working definition of digital library’’, available at: www.clir.org/diglib/dldefinition.htm (accessed 1 December 2007). Greenberg, J. and Me´ndez, E. (2007a), ‘‘Introduction: toward a more library-like web via semantic knitting’’, Cataloging & Classification Quarterly, Vol. 43 No. 3-4, pp. 1-8. Greenberg, J. and Me´ndez, E. (Eds) (2007b), Knitting the Semantic Web, The Haworth Information Press, Binghmaton, NY (published simultaneously as Cataloging & Classification Quaterly, Vol. 43 No. 3-4). Legg, C. (2007), ‘‘Ontologies on the Semantic Web’’, in Cronin, B. (Ed.), Annual Review of Information Science and Technology, Vol. 41, Information Today Inc., New Jersey, NJ, pp. 407-51. Miller, P. (2005), ‘‘Web 2.0: building the new library’’, Ariadne, No 45, available at: www.dlib.org/ dlib/april06/miller/04miller.html (accessed 1 December 2007). Miller, P. (2006), ‘‘Coming together around Library 2.0: a focus for discussion and a call to arms’’, D-Lib Magazine, Vol. 12 No. 4, available at: www.dlib.org/dlib/april06/miller/04miller.html (accessed 1 December 2007). Prasad, A.R.D. and Madalli, D.P. (Eds) (2007), Proceedings of the International Conference on Semantic Web and Digital Libraries (ICSD-2007), 21-23 February 2007, Documentation Research and Training Centre, Bangalore. Corresponding author George Macgregor can be contacted at: [email protected]

To purchase reprints of this article please e-mail: [email protected] Or visit our web site for further details: www.emeraldinsight.com/reprints

Guest editorial

177

Antaeus

Heracles defeating Antaeus. Public domain image: from the old Swedish encyclopedia Nordisk familjebok *The title of the ‘‘Antaeus’’ column derives from the name of the mythical giant, Antaeus or Antaios. The son of Gaia (whose name means ‘‘land’’ or ‘‘earth’’), Antaeus was undefeatable in combat so long as he remained in contact with the earth. Once grounded by contact with the soil, he vanquished all opponents. However, in order to disempower Antaeus, Heracles simply lifted him from the earth, overcoming him totally. Thus, many times through the centuries, Antaeus has been used as a symbolic figure showing how any human aspiration must remain grounded in order to succeed. LIS research must therefore retain its contact with the ‘‘ground’’ of everyday practice in order to fulfil its potential as a sophisticated research discipline – it must remain empowered by its relevance to practitioners.

The current issue and full text archive of this journal is available at www.emeraldinsight.com/0024-2535.htm

ANTAEUS

The practitioner librarian and the semantic web Nicholas Joint Centre for Digital Library Research/Andersonian Library, University of Strathclyde, Glasgow, UK Abstract

The practitioner librarian and the semantic web 179 Received 16 December 2007 Reviewed 18 December 2007 Accepted 21 December 2007

Purpose –To describe and evoke the potential impact of semantic web systems at the level of library practice. Design/methodology/approach – A general outline of some of the broad issues associated with the semantic web, together with a brief, simple explanation of basic semantic web procedures with some examples of specific practical outcomes of semantic web development. Findings – That the semantic web is of central relevance to contemporary LIS practitioners, whose involvement in its development is necessary in order to determine what will be the true benefits of this form of information service innovation. Research limitations/implications – Since much of the initial discussion of this topic has been developmental and futuristic, applied practitioner-oriented research is required to ground these discussions in a firm bedrock of applications. Practical implications – semantic web technologies are of great practical relevance to areas of LIS practice such as digital repository development and open access services. Originality/value – The paper attempts to bridge the gap between the abstractions of theoretical writing in this area and the concerns of the working library professional. Keywords Libraries, Digital libraries, Worldwide web Paper type Viewpoint

Introduction Great steps forward often happen when two factors merge: someone may have a great idea, but someone else needs to have an application for that great idea for real progress to be made. When the great idea and the ‘‘killer application’’ meet, we then witness one of those exciting moments of creative advancement, which, in retrospect seem to be clear, obvious and laudable by all. With the benefit of hindsight we applaud and somehow assume that we knew something was a good thing all along. At the everyday level of library practice, just now the ‘‘Semantic Web’’ looks like a great idea which is still awaiting its big opportunity for a wide-ranging relevant application. It is like HTML when it was just a nice way for physics boffins to exchange information at CERN, or the Internet when it was ARPANET, merely a robust, highly distributed way for the US defence establishment to network electronic data in the face of catastrophic attack. These were complex and clever things, but nobody knew that much about them, and if anyone did, they appeared too highly specialised to be relevant to everyday information practice. Obviously, since the early days of CERN and ARPANET, hypertext and the Internet have proven to be transformational throughout all levels of science and society – and certainly supremely relevant to all information professionals and information users! Not only have they proven themselves as great ideas with a wealth of applications. But a third factor also applies: the essence of them is easily understood. This short paper will assume that the Semantic Web is – as most would acknowledge – a sophisticated and exciting concept. But it will also try to show that it is genuinely of

Library Review Vol. 57 No. 3, 2008 pp. 178-186 # Emerald Group Publishing Limited 0024-2535 DOI 10.1108/00242530810865466

LR 57,3

180

enormous potential relevance to information practice. Moreover, a third goal will be to show how the concept of the Semantic Web possesses that supreme feature of any great advance – its essence is indeed really quite simple (though descriptions of Semantic Web applications are often highly abstract and intimidatingly complex, even hermetic). This does not mean that one can definitively and straightforwardly describe what the future of the Semantic Web will be, nor absolutely ‘‘prove’’ that it will be a transforming innovation in information delivery. However, its features are such that it needs to be understood by information professionals who may then make informed decisions as to how to modify their information practice in the light of their understanding of it. Principles and origins The opening three paragraphs of the Semantic Web Activity home page[1] give a good sense of the elegant simplicity of the Semantic Web concept. To paraphrase and summarise them: .

If I can see my photographs, and I can see my appointments in a calendar on the web, why can not I see my photos in a calendar to see what I was doing when I took them?

.

. . . because data is controlled by applications, and each application keeps it to itself.

.

So, to overcome this, the Semantic Web looks at common formats for integration and combination of data drawn from diverse sources, not just the interchange of documents; and at the language for recording how the data relates to real world objects.

Another useful ‘‘librarian-friendly’’ way to generate a feel for Semantic Web issues is to think back to the first library experiences of HTML authorship and web page generation. Librarians became absorbed in HTML applications because of immediate professional need, not because hypertext was intrinsically interesting (although once hooked on HTML, hypertext soon grew to absorb librarians’ attention and even excite their enthusiasm). This original immediate need was two-fold: first, offering web pages was a much better way than paper to distribute textual descriptions about libraries, their collections and how to use them; and secondly, the information resources that librarians provided mutated into web resources which needed a coherent portal of links to make them usable. The typical library web site is thus a task-oriented collection of HTML pages, responding to two user task requirements: ‘‘I want to read passive text about using local information resources’’; and, ‘‘I want to access and use online information resources interactively’’. So librarians to some extent stumbled into the information science of human-computer-interaction and interface design, and became pretty good at applying these disciplines to good practical effect. Logically, therefore, librarians will appreciate the strength of the original HTMLbased structure of the web as a medium for the simple interchange of documents, but also the concomitant frustrations inherent in its HTML construction. HTML is a format-driven system which marks up data items in such a way as to give them a certain presentational style that can be readily displayed over the network. As such, it is not designed to express data structures within a given web page, nor, more

ambitiously, is it a system that readily expresses semantic relationships between related data that is shared by HTML resources distributed across the world wide web. As librarians know, the limitations of HTML as a medium for the expression of data structures at the page level gave rise to a simple but elegant solution via the use of XML and cascading style sheets (CSS). The style sheet became the repository of stylistic elements, while the underlying page file itself became the repository of content, merged at the user level in the browser. The separation of stylistic elements from semantic elements meant that the data structure of a web page became exposed when the formal syntax of the stylistic elements was abstracted away from the page. Put more simply, the discipline of using CSS made librarians aware of the stupidity of making (say) the three most important top level concepts in a web page look different from each other; or making them even less noticeable than subsidiary text strings by using a skewed mark-up format (e.g. illogical heading tags, or font and size commands that confusingly mimicked heading functions). In early HTML, where mark-up resided casually in and around the text content of the web page, these mistakes were perilously easy to commit. In XML/CSS, the discipline of this more evolved mark-up language makes these errors much harder, by revealing the latent data structure of the web page, albeit in a rigidly hierarchal fashion that is logical only within the page itself or at best, by analogy, across equally formally constrained pages sharing the same data structure within the same website. The data web In writing about the Semantic Web, Tim Berners-Lee has expressed his vision of the original world-side web as being a ‘‘data web’’ rather than the current, less usable web[2]. Effectively, the need to express latent ‘‘data structures’’ from the semantic soup of crude HTML is a naturally evolving process which, as we can see, has involved librarians for some time as part of their normal professional practice. As outlined above, the simple steps forward which have already been made in growing the data web show how ‘‘proto-Semantic Web’’ principles have suffused LIS web practice almost unconsciously from the start of web-based digital library work, with immediate benefits. For example, as a result of the improving XML/CSS data web standard, librarians are able to create websites which are fully interoperable with assistive technology, and which permit different stylistic expressions of the same core data to suit visual and reading impairments. This is only possible because of a commitment to a disciplined semantic distinction expressed through a formal language. Similarly, a more widely applicable benefit is that data preservation for library web collections is facilitated by separating persistent content from changing mark-up. It is only common sense therefore for librarians to follow the logical evolution of the data web, since there have been so many practical benefits already from attempts to expose web-based data structures more cleanly, logically and transparently. This is a journey which we have already started – though we may not have realised it – so it would be eccentric not to follow the ride to its conclusion, when its early stages have gone so well for us! The real ‘‘data web’’ is one in which an abstract system encodes the relationships between data across the broad web (rather than at the page level). The means which have been created to describe the resources between which these relationships exist is the resource description framework (RDF)[3]. If ‘‘resource description’’ sounds rather like cataloguing or bibliographic description, that is fair enough, since the RDF enables

The practitioner librarian and the semantic web 181

LR 57,3

you to describe any resource that is locatable on the web by predicating qualities to that resource. Thus, RDF is based on subject-predicate-object statements, or ‘‘triples’’: (1) a subject (an item with a URL) has a (2) predicate (e.g. title), which has

182

(3) value, a given text string (the words of the title), which text string is the object predicated to the subject. In turn, each component of the triple is customarily identified by a uniform resource identifier (URI) – for a Dublin Core title predicate, the URI would be . RDF is thus a ‘‘mark-up’’ language into which library-like cataloguing systems such as Dublin Core fit well, since the Dublin Core metadata elements provide a consistent way of saying specific things about a resource. And beyond the fundamental act of description, the Semantic Web works in terms of the meaningful relationships between the resources; something which structured description facilitates. These relationships are the result of using consistent vocabularies or ontologies, thus creating a set of data that is as ordered as a database but distributed rather than existing in any centralised sense. There are quite a few overviews and introductions to Semantic Web technologies and tools available (e.g. Greenberg and Me´ndez, 2007). These explain the principles more fully, but hopefully these basic generalisations give some sense of how this highly sophisticated view of a fully achieved ‘‘data web’’ occupies a position on a continuum that began with earlier ‘‘data web’’ principles; principles that have been important in practice for practitioner librarians. The implication is that the tangible, initial benefits of library involvement in the vestigial data structures of the early web were merely the ‘‘low hanging fruit’’ of the web’s data richness, and thus give only a taste of the benefits that would come as a result of fully committed librarian involvement in the Semantic Web project. Relevance in practice In trying to sell to the practitioner the benefits of a mature Semantic Web, Eric Miller has summed up the role of libraries under four headings[4]: (1) exposing collections – use Semantic Web technologies to make content available; (2) web’ifying thesaurus/mappings/services; (3) sharing lessons learned; (4) persistence. What this means in everyday library work needs careful consideration however. Skills, tools and techniques The idea of sharing library expertise and adapting traditional library taxonomies and schemes to the Semantic Web is interesting, but slightly quixotic. Librarians have long seen the amorphous web as a type of failed library, one desperately in need of the sorts of disciplines and techniques that the profession has long used in the hard copy information world. In this view, librarians are the knights in shining armour coming to the rescue of the web. But helping tame that amorphous void is not necessarily such an

attractive proposition when there are plenty of problems lurking in one’s own backyard. Yes, library skills honed in the non-web world can be donated to the world beyond. But if the inhabitants of the non-library world want to use these techniques, they are welcome to do so, and they can manage the process for themselves. The non-Semantic Web world of library print and electronic collections still seems quite distinct and self-contained, regardless of the transferability of its tools and techniques. Outsiders can look at it and learn from its spin-off without our involvement. The library as closed application Let us not challenge this complacent and limited point of view[5] and move on to say (albeit with gritted teeth) that the idea of exposing collections more effectively could be a more potent argument. This is still a problematic position, in that library professionals do not need to expose collections, other than to a tightly defined readership: for example, the typical hybrid library consists of highly organised, already semantically coherent suites of electronic services which are self-contained and which are used in tandem with a print collection that is eminently searchable via an online public access catalogue. The need to better expose these collections is limited by the fact that hard copy collections are geographically and physically circumscribed so that exposing their metadata more widely is not a top priority (hence, there is no practitioner demand for a RDF-enabled Semantic Web opac). And electronic collections, especially full text ones, are again limited by licence and digital rights restrictions to members of the home institution. Librarians spend more time using authentication systems to restrict exposure of these collections than in dreaming of making these systems interoperable with an outside world that does not pay their wages. If the Semantic Web project rages against a status quo in which data is controlled by applications, because each application keeps it to itself, then ‘‘the library’’ is in one sense the ultimate self-contained application! The library as an open medium of institutional data exposure However, library professionals have already had their first experiences of the need to expose digital content as widely as possible via the construction of library websites made up of library hand outs and links. This trend has become ever more important, for example through the construction of ambitious ‘‘heritage’’ library websites, where digitised content derived from print originals which the parent institution either owns, or which is out of copyright, is made available to the entire web community. And again the growth of open access repositories, into which the born-digital content originated by an institution’s content-generators is deposited, gives enormous momentum to Semantic Web developments at the heart of library practice. Firstly, bodies that fund research are committed to maximising open access and will not look kindly on institutions who do not share this publicly funded research as widely as possible over the most appropriate platforms in as flexible and configurable a format as possible. Libraries increasingly support a range of digital repositories, with open access research repositories being a prime driver of such developments. Librarians will thus inevitably need to look to Semantic Web developments as a prime means of presenting data for maximum benefit to the world research community.

The practitioner librarian and the semantic web 183

LR 57,3

184

Secondly, another powerful impulse behind open access repositories is the need to create an effective system of metrics for research evaluation. Again, the key principal underlying metrics-based evaluation is achieving the best possible exposure of institutional content to maximise its use and increase the impact and profile of institutional research materials. Semantic Web approaches are compelling in their ability to order and make this type of content best available to the widest possible community of interest. At present, LIS practitioners are poring over user statistics for their repositories, deliberating on strategies for raising the profile and usability of this data on the global web. On the one hand, they are looking for evidence that they are fulfilling their open access aspirations; on the other, they know that the more research is read and cited, the greater the financial reward given to the institutions whose authors created that research (not all arguments for the Semantic Web as a tool of open access have to be disinterested!). It is quite illogical therefore for the practitioner community not to share Miller’s positive view on how libraries should use Semantic Web technologies to expose collections and to make content available: this is what we do now. It is less a vision of the future than a statement of present fact. So we must learn to fulfil that role as effectively as possible. Conclusion Of course, none of this should be taken as proof that the Semantic Web project is the only significant way forward for the future of well-organised, networked information. The indisputably evangelical tone of many statements about the semantic development of the original web tends to inspire many highly intelligent critics to bite back (for example, Shirky, 2003). This equal and opposite reaction is well reasoned and intelligent: many ideas are coherent and logical, but do not transform the world of every day applications. Cars are powered by internal combustion engines, not by external combustion engines, although everyone knows that the Stirling engine is in itself a brilliant idea. Functional programming is a fascinating, mathematical way of writing computer programmes, but in practice it is nothing like as good as mainstream programming methods. These are brilliant, compelling scientific systems, but irrelevant to the world in which we only want something that works. But it is a fact that library practice is at a cross roads, caught between two views of the information world. One is dualist and traditional, the other is single and visionary. It is at least a logical hypothesis that the transformation from one order to the other can be engineered by Semantic Web technologies. This is an experimental hypothesis that should be examined empirically and seen though to a full conclusion. In this experiment, we can see that there is a traditional mindset which sees existing library systems as a ‘‘given’’, in which there is an acceptable and necessary separation between what is properly the library’s business from that other world of non-library data. This viewpoint appears happy with the library as a closed application, distinct from the open world of the data web. The more forward-thinking viewpoint sees the world of the data web already infiltrating the self-containment of the library and rendering its data transparent. It is in fact the express hope of the library community that the open access model blends into the self-contained world of the closed access hybrid library to create a single information world. We as a profession are committed to phasing out one model in favour of the other[6]. What we lack is an ‘‘engine’’ to act as the dynamic motor of this

change. We need to fully explore the challenging question: is the Semantic Web the engine of this change? Even if the eclipse of one information model by another is for now an aspiration not an achievement, there is no reason why the existing ‘‘library as closed application’’ model of information provision cannot take on more and more of the attributes of the Semantic Web and bring on board many of its information-enriching tools and techniques. Any superficial belief that the borrowing of our metadata schemes and taxonomies is a one-way street is deeply misinformed: the reason why we should actively participate in this sharing of skills and tools is because we could well be shaping our own future information landscape, not just helping form a better but quite discrete non-library WorldWideWeb. The Semantic Web is thus truly a subject with which the responsible reflective practitioner must engage. What their final verdict might be is open to debate – but participation in the debate is an exciting and stimulating challenge for the twenty-first century librarian. Notes 1. W3C Semantic Web activity, available at: www.w3.org/2001/sw/ (accessed 11 December 2007). 2. ‘‘. . . now people understand that the Semantic Web is the Data Web. I think we could have called it the Data Web. It would have been simpler’’. Q&A with Tim Berners-Lee: the inventor of the Web explains how the new Semantic Web could have profound effects on the growth of knowledge and innovation. BusinessWeek Special Report: April 9, 2007, available at: www.businessweek.com/technology/content/apr2007/tc20070409_961951.htm (accessed 11 December 2007). 3. W3C.Technology & Society Domain: Resource Description Framework (RDF), available at: www.w3.org/RDF/ (accessed 11 December 2007). 4. The Semantic Web and Digital Libraries. Eric Miller, W3C. DC 2004/SILF 2004. Shanghai Library, Shanghai, China. 2004-10-13, available at: dc2004.library.sh.cn/ english/prog/ppt/talk.ppt (accessed 11 December 2007). 5. A narrow and partial view of participation in developmental work in this area is arguably in breach of core librarian professional ethics, given the comprehensive commitment of our professional bodies to engagement in the formation of a fully achieved information society. Our professional mission at national level is variously: ‘‘to enhance learning and ensure access to information for all’’, American Library Association, Available at: www.ala.org/ala/ourassociation/governingdocs/policymanual/mission.htm (accessed 11 December 2007). ‘‘To set, maintain, monitor and promote standards of excellence in the creation, management, exploitation and sharing of information and knowledge resources’’ (CILIP), available at: www.cilip.org.uk/aboutcilip/charterMissionGoals (accessed 11 December 2007). . . . leading to this overarching international commitment: ‘‘IFLA advocates a global information commons through which all people will be enabled to seek and impart information’’, available at: www.ifla.org/III/wsis/wsis-24Feb05.html (accessed 11 December 2007). These collective commitments are incompatible with disengagement from the semantic web, which offers potentially the best model of ‘‘a global information commons’’ that has yet been proposed. 6. (a) IFLA Statement on Open Access to Scholarly Literature and Research Documentation, available at: www.ifla.org/V/cdoc/open-access04.html (accessed 11 December 2007). (b) Charles Bailey Jr., Open Access Bibliography. Open Access Statements: available at: www.escholarlypub.com/oab/2statements.htm#2.1 (accessed 11 December 2007).

The practitioner librarian and the semantic web 185

LR 57,3

186

References Greenberg, J. and Me´ndez, E. (Eds) (2007), Knitting the Semantic Web, Vol. 43 No. 3-4, Haworth Information Press, New York, NY (simultaneously published as Cataloging & Classification Quarterly). Shirky, C. (2003), ‘‘The Semantic Web, syllogism, and worldview’’, Networks, Economics, and Culture, November 7, available at: www.shirky.com/writings/semantic_syllogism.html (accessed 11 December 2007). Corresponding author Nicholas Joint can be contacted at: [email protected]

To purchase reprints of this article please e-mail: [email protected] Or visit our web site for further details: www.emeraldinsight.com/reprints

The current issue and full text archive of this journal is available at www.emeraldinsight.com/0024-2535.htm

Integrated access to cultural heritage resources through representation and alignment of controlled vocabularies

Cultural heritage resources

Antoine Isaac and Stefan Schlobach

Received 22 October 2007 Reviewed 1 November 2007 Accepted 16 November 2007

Department of Computer Science, Vrije Universiteit Amsterdam, Amsterdam, Noord-Holland, The Netherlands

187

Henk Matthezing Koninklijk Bibliotheek, The Hague, The Netherlands, and

Claus Zinn Max-Planck Institute for Psycholinguistics, Nijmegen, The Netherlands Abstract Purpose – To show how semantic web techniques can help address semantic interoperability issues in the broad cultural heritage domain, allowing users an integrated and seamless access to heterogeneous collections. Design/methodology/approach – This paper presents the heterogeneity problems to be solved. It introduces semantic web techniques that can help in solving them, focusing on the representation of controlled vocabularies and their semantic alignment. It gives pointers to some previous projects and experiments that have tried to address the problems discussed. Findings – Semantic web research provides practical technical and methodological approaches to tackle the different issues. Two contributions of interest are the simple knowledge organisation system model and automatic vocabulary alignment methods and tools. These contributions were demonstrated to be usable for enabling semantic search and navigation across collections. Research limitations/implications – The research aims at designing different representation and alignment methods for solving interoperability problems in the context of controlled subject vocabularies. Given the variety and technical richness of current research in the semantic web field, it is impossible to provide an in-depth account or an exhaustive list of references. Every aspect of the paper is, however, given one or several pointers for further reading. Originality/value – This article provides a general and practical introduction to relevant semantic web techniques. It is of specific value for the practitioners in the cultural heritage and digital library domains who are interested in applying these methods in practice. Keywords Worldwide web, Archives management, Digital libraries Paper type Conceptual paper

Introduction: the semantic interoperability problem In the digital age, cultural heritage (CH) institutions have the opportunity to, and face the challenge of, using the World Wide Web to make accessible the digital artefacts of their collections, together with their metadata. Web-based access to digitised images and their descriptions, at anytime from anywhere, lowers the barriers for access to information resources. Once there is digital access to the content of museums, libraries and archives, there is also the tremendous opportunity to merge collections from This paper is based on a talk given at ‘‘Information Access for the Global Community, An International Seminar on the Universal Decimal Classification’’ held on 4-5 June 2007 in The Hague, The Netherlands. An abstract of this talk will be published in Extensions and Corrections to the UDC, an annual publication of the UDC consortium.

Library Review Vol. 57 No. 3, 2008 pp. 187-199 # Emerald Group Publishing Limited 0024-2535 DOI 10.1108/00242530810865475

LR 57,3

188

Figure 1. Semantic heterogeneity hampers collection access

different locations into virtual, federated institutions, thus increasing access across collections and institutional boundaries. In stark contrast to the vast amount of existing digital resources on the World Wide Web, CH assets from libraries, museums and archives are very well described. Over many generations, librarians, curators and archivists have developed knowledge organisation systems (KOSs) – controlled vocabularies such as thesauri, classification schemes and ontologies – to organise and manage their collections. The organisation and access to CH, and the human capacity to deal with information and knowledge, is a valuable achievement in itself. It helps us to grasp our past and present, and this understanding must be exploited to facilitate its access at a grander scale. The move toward cross-institutional CH portals is well under way, as, for instance, The European Library[1] and the memory of the Netherlands[2] testify. In this paper, we describe how CH expertise can be combined with knowledge and technology from the Semantic Web (SW) community to deliver portals that provide a seamless and unified access to different collections via semantic search and navigation. Figure 1 illustrates the problem that needs to be solved in a networked environment. Consider two collections, each of which is indexed by its dedicated knowledge organisation system. Instead of using one single conceptual vocabulary for querying or browsing the objects of both collections simultaneously, users are expected and required to use the terminology of the first KOS to identify objects of the first collection, and the second KOS to identify those of the second collection. We say that these two KOSs are not interoperable at the semantic level. In the given example, when searching for objects showing a ‘‘Madonna’’ one will only retrieve objects that were indexed using this specific subject description (the statue in the upper right); one will not find the manuscript illumination (in the lower right) that was indexed as ‘‘Virgin Mary’’, which is clearly a conceptually similar subject description, but stems from another controlled vocabulary. Not taking care of the semantic heterogeneity of their respective KOSs when merging collections clearly hampers the ease of accessibility. The burden of search is indeed transferred to users who then need to perform two well-formulated queries

(using the respective correct terminology) to obtain the desired objects from the two collections. Two heterogeneity problems must be solved to enhance the interoperability of controlled vocabularies and, hence, of the systems and collections that use them: (1) Representational heterogeneity: vocabularies often come in different formats; some will be encoded in XML while others will come as plain text. Moreover, the models guiding their design might not be directly compatible. They might mirror different general information needs (e.g. thesauri contain ‘‘terms’’ while classification schemes contain ‘‘classes’’), and different KOS might have different kinds of notes and labels attached to conceptual entities, for instance.

Cultural heritage resources

189

(2) Conceptual heterogeneity: any two vocabularies will usually contain concepts that have identical or similar meanings but different labels or names (e.g. like ‘‘Virgin Mary’’ and ‘‘Madonna’’). Also, there will be concepts that are more general than others (e.g. like ‘‘Mother’’ and ‘‘Virgin Mary’’). Such similarity and subsumption links have to be determined and exploited so that an integrated system can provide users with seamless access to joint content described by several vocabularies. In this paper, we show how these two problems can be addressed using techniques that are currently being investigated in the Semantic Web research domain. In section 2, we describe the basic elements of the Semantic Web infrastructure, and illustrate how the simple knowledge organisation systems (SKOS) standard model can be used to represent different vocabularies KOSs homogeneously. In section 3, we show how the representation of the different vocabularies, then commonly represented in the SKOS format, can be semantically aligned to enable a semantic integration of different collections. Finally, in section 4, we demonstrate how we solved a real-life problem with a combination of SW techniques, and briefly describe the resulting prototype. Semantic Web techniques and controlled vocabulary representation The Semantic Web (Berners-Lee et al., 2001) is a proposed extension of the existing web, where information found on the web is augmented with machine-accessible knowledge[3]. The basic building blocks of the Semantic Web, as introduced by the resource description format (RDF)[4], are resources which denote any element that can be identified on (or even outside) the web. These resources are described by three-part statements that link them together. Each statement has a subject resource which is linked to an object resource via a property resource. Together, several such triplets form a graph, such as the one represented in Figure 2[5]. These graphs can contain:

Figure 2. A Semantic Web RDF graph[6]

LR 57,3

190

(1) Factual knowledge: the third paragraph of the described document is about ‘‘Amsterdam’’; the type of the described document is ‘‘Article’’; and the selected paragraph ‘‘par3’’ is part of a (larger) file called ‘‘file1’’. (2) Ontological knowledge: the Semantic Web is concerned about the way resources can be grouped in conceptual classes. These classes are introduced in ontologies that contain formally expressed knowledge about them. Here, ‘‘Article’’ is a class more specific than (or a subclass of) ‘‘Document’’. The information contained in ontologies is important, since it provides material for automated reasoning on the resources which populate the classes. For example, from the information found in Figure 2 for ‘‘file1’’, ‘‘Article’’ and ‘‘Document’’, an automated reasoning engine can infer that ‘‘file1’’ is also an instance of the ‘‘Document’’ class, which will yield more answers for queries containing ‘‘Document’’. It should be noted that the RDF framework is designed to allow different sources of knowledge to co-exist with each other, inhabiting the same space. This means that Semantic Web data can merge and operate with resources coming from different information spaces. In our example, the objects and links in Figure 2 come from different namespaces, either user-defined (myVoc1:, myVoc2:) or predefined (RDF:). The resource ‘‘Amsterdam’’ in myVoc2: may indeed refer to the capital of The Netherlands (as the RDF graphs in which it occurs would show), while some other resource with the same name, but from a different vocabulary space, may refer to a city in the state of New York, USA. In any case, both resources stem from different name spaces and can both inhabit different contexts, further defining and constraining their intended meaning. RDF ‘‘triples’’ are the basic building blocks for translating KOS into a homogeneous format. Also, in order to mirror a KOS’ modelling elements (e.g. the ‘‘broader than’’, or ‘‘narrower than’’ relation types of thesauri), additional constructs are necessary. RDFSchema (Brickley and Guha, 2004), in short RDF-S, is a simple representation language that allows users to define their models, introducing different types for RDF resources and links. One can also express, for instance, that the source and target of a relation are of a specific type, e.g. that the relation type ‘‘has painted’’ requires a subject of type ‘‘painter’’ (or ‘‘artist’’), and an object of type ‘‘painting’’ (or ‘‘drawing’’). The current standard web ontology language is called OWL (McGuinness and Harmelen, 2004). This formal language is more expressive than RDF-S, allowing users to define a variety of different properties of classes and relations between them. A more detailed discussion of OWL is beyond the scope of this paper. To support experts in converting their KOSs into the RDF-based formats, but also to facilitate the future exchange of such formats, the World Wide Web Consortium (W3C) has initiated the development of SKOS[7], a standard model that allows CH practitioners (and other terminologists) to homogenously represent the basic features of KOSs. SKOS introduces a set of constructs for RDF, which mainly allow for the description of concepts and concept schemes (Miles and Brickley, 2005). Concept description SKOS has chosen a concept-based approach for the representation of controlled vocabularies. As opposed to a term-based approach, where terms from natural language are the first-order elements of a KOS, SKOS describes abstract concepts that may have a different materialisation in language (lexicalisations). SKOS introduces a special construct skos:Concept[8] to properly characterise the (web) resources that

denote such KOS elements. To further specify these conceptual resources, SKOS features: (1) Labelling properties, e.g. skos:prefLabel and skos:altLabel, to link a concept to the terms that represent it in language. The prefLabel value is a non-ambiguous term that uniquely identifies the concept, and can be used as a descriptor in an indexing system. The term altLabel is used to introduce alternative entries, such as synonyms, abbreviations, and so forth. SKOS allows concepts to be linked to prefLabels and altLabels in different languages. SKOS concepts can thus be used seamlessly in multilingual environments. (2) Semantic properties are used to represent the structural relationships between concepts, which are usually at the core of controlled vocabularies like thesauri. The construct skos:broader denotes the generalisation link (BT in standard thesauri), while skos:narrower denotes its reciprocal link (NT), and skos:related the associative relationship (RT). (3) Documentation properties. Often, informal documentation plays an important role in a KOS. SKOS introduces explanatory notes – skos:scopeNote, skos:definition, skos:example – and management notes – skos:changeNote, skos:historyNote, etc. Concept scheme description A KOS as a whole also has to be represented and described. SKOS coins a skos:ConceptScheme construct for this. It also introduces specific properties to represent the links between different KOSs and the concepts they contain. The term skos:inScheme asserts that a given concept is part of a given concept scheme, while skos:hasTopConcept states that a KOS contains a concept as the root of (one of) its constituent hierarchical tree(s) (i.e. a concept without a broader concept). Conversion from a KOS native representation to SKOS RDF data requires the analysis of the original model of the KOS, and the linking of the elements of this model to the SKOS ones that fit them most (Assem et al., 2005). One can, for instance, decide to represent a ‘‘class’’ in a classification scheme as a resource of type skos:Concept. Based on such a specification, it is then possible to implement an appropriate conversion program – e.g. an XSL stylesheet when the vocabulary is natively encoded in XML – to automatically convert the initial representation to a SKOS one. As an example, a subject 11F coming from the Iconclass concept scheme[9], ‘‘the Virgin Mary’’, identified by the (as yet fictive) resource http://www.iconclass.nl/s_11F, could be partly represented by the graph in Figure 3. Vocabulary alignment as a solution to the interoperability problem Having unified and linkable representations of the concepts contained in different collections’ vocabularies helps managing them in a single framework. However, this is not sufficient for solving the semantic interoperability problem. One still has to determine semantic similarity links between the elements of the different vocabularies – to align[10] them (Doerr, 2001). Figure 4 illustrates that if a search engine ‘‘knew’’ that a SKOS concept C from a thesaurus T1 is semantically equivalent to a SKOS concept D from thesaurus T2, then it could return all the objects that were indexed against D for a query for objects described using C. The objective is therefore to align as many concepts of one thesaurus to their semantic equivalents in the other thesaurus. Where such equivalency cannot be established, it may be possible to establish links between concepts of one thesaurus and

Cultural heritage resources

191

LR 57,3

192

Figure 3. A SKOS graph partly representing the Iconclass subject 11F

Figure 4. Using vocabulary alignment for integrated access to different collections

concepts of the second thesaurus that are either more specific or more general, and to exploit such ‘‘narrower than’’ and ‘‘broader than’’ relations for query processing. Such an approach has been investigated for subject vocabularies in projects such as HILT (Macgregor et al., 2007). The alignment of these vocabularies is however a labourintensive task that requires considerable expertise in the concerned thesauri. Manual alignment has been approached by several projects, notably, CARMEN (Krause, 2003), Renardus (Day et al., 2005), KoMoHe[11], AOS (Liang and Sini, 2006) or the ongoing CRISSCROSS[12], MACS[13] (Landry, 2004) and MSAC (Balikova, 2005). These projects have yielded very interesting results such as the development of tools to support manual alignment, the deployment of search engines that exploit resulting alignments, and the contribution of initial methodological ideas. However, they also demonstrated the complexity, difficulty and cost of manually aligning large vocabularies (usually containing many thousand concepts) in realistically-sized collections and settings.

Given that manual labour is expensive and that vocabularies evolve over time, it is clear that the construction and maintenance of alignment constitutes an important issue that needs to be addressed. There is a need for developing advanced, computerbased tools that can identify candidate mappings between two vocabularies, and that can then propose them to the human expert for consideration. Alignment would thus become a semi-automatic task where thesaurus experts’ work would be assisted, and where the integration of collections would become more cost-efficient. Recently, the Semantic Web community has produced alignment tools that address the specific problem of formal ontology matching (Shvaiko and Euzenat, 2005). However, the techniques they employ and the goals they advertise make them deployable in a more general context, including thesauri and other similar KOSs. Although most of the existing ontology alignment tools rely on sophisticated methods (Euzenat and Shvaiko, 2007), they can be classified and described by the basic techniques they build upon and the different sources of information they exploit: the lexical information attached to the concepts of the vocabularies, the structure of vocabularies, the collection objects described by vocabularies, or other (external) knowledge sources. Lexical alignment techniques In these techniques the lexical materialisations of the concepts are compared to each other. If a significant similarity is found, then we can establish a semantic link between the concerned concepts. A straightforward example is when two concepts have the same label. But one can also search for string inclusion patterns or more complex techniques relying for instance on lemmatisers – getting normalised forms of labels, e.g. ‘‘tree’’ for ‘‘trees’’ – and syntactical analysis tools. A concept labelled ‘‘(map of) the North Pole’’ can be detected as more specific than a concept ‘‘Charts, maps’’. These lexical methods exploit the preferred labels of concepts, but they can also turn to their lexical variants or their associated definitions and scope notes. Clearly, such approaches encounter the same problems as humans when dealing with words taken out of context. Polysemy and homonymy, for instance, are common sources of errors. This has to be compensated with contextual information. Structural alignment techniques The first kind of context is provided by the vocabulary itself, as it contains hierarchical and associative links between concepts. These links, especially those concerning hierarchical generalisation and specialisation, are useful to constrain the natural interpretation of a concept: ‘‘bank’’ will be understood differently if it is a narrower term of ‘‘finance’’ or ‘‘geography’’. Some tools will analyse this semantic context, either to check similarities obtained by other techniques or to derive new similarities from existing ones. If two concepts from different vocabularies are semantically equivalent, this equivalence will positively influence the alignment tool when it will examine the children of these concepts to find similarities between them. Extensional alignment techniques The second kind of context comes from the actual usage of the concepts in real-life applications. For instance, a class from a classification scheme will be used to categorise a number of objects in a collection (e.g. books). Accessing this information will provide an ‘‘extensional’’ characterisation of the class’ intended meaning – akin to its literary warrant. When documents are described using two different

Cultural heritage resources

193

LR 57,3

194

Figure 5. Using object-level information to align vocabularies

vocabularies[14], statistical techniques can be employed to compare the sets of documents described by the concepts from these vocabularies (Figure 5). A high degree of overlap between these sets will yield a high similarity between corresponding concepts. Several such techniques have already been experimented in the KOS field, as in (Zhang, 2006) or (Isaac et al., 2007). Background knowledge-based alignment techniques A final group of alignment methods rely on knowledge sources that are external to the application and the vocabularies being considered. Sources of different kinds can be used, for instance general-purpose ontologies like CYC[15] or semantic networks like Wordnet (Miller, 1995). These sources can contribute KOS-external knowledge to compensate for the lack of KOS-internal lexical or structural information. For example, a concept ‘‘calendar’’ from one thesaurus can be aligned to the more general concept ‘‘publication’’ from another thesaurus, using the hypernymy relation that holds between the two corresponding terms in Wordnet (Figure 6). Integrated collection access: an example To illustrate the potential of the described technology, we used it for creating integrated access to two collections belonging to two different Dutch CH institutions,

Figure 6. Using background knowledge to align vocabularies

the Rijksmuseum, and the National Library of the Netherlands (Gendt et al., 2006). The manuscripts collection contains 10,000 medieval illuminations, which are annotated by subject indices describing the content of the image. These indices come from the Iconclass classification scheme, a vocabulary of 25,000 elements designed for iconographical analysis. The masterpieces collection contains 700 objects such as paintings and sculptures, and its subjects are indexed using the Amsterdam Rijksmuseum Inter Actief (ARIA) ‘‘catalogue’’, a vocabulary conceived mainly as a resource for hierarchical browsing. Both vocabularies were translated into SKOS, and mappings between them were calculated with existing state-of-the art mapping tools, namely, Falcon (Jian et al., 2005) and S-Match (Giunchiglia et al., 2005). Falcon uses a mixture of lexical and structural techniques. In addition to lexical techniques, S-Match uses Wordnet as background knowledge, and exploits ‘‘semantic reasoning’’ using a logical interpretation of the concepts based on the structure of the vocabularies. We implemented a faceted browser, in which the mappings and the vocabularies’ Semantic Web representations are exploited to provide integrated assess to the collections, offering three different views: single, combined, and merged view. The single view presents the integrated collections from the perspective of just one of the vocabularies. In the screen capture (Figure 7), the first four pictures come from the Rijksmuseum, the others are illuminated manuscripts. Browsing is done solely using the ARIA catalogue (i.e. these illuminations have been selected exploiting the mapping between the currently selected ARIA concept ‘‘Animal Pieces’’ and the Iconclass concept ‘‘25F:animals’’). The combined view provides simultaneous access to the collections through their respective vocabularies in parallel. This allows us to browse through the integrated collections as if it was a single collection indexed against two vocabularies. In Figure 8, we made a subject refinement to ARIA ‘‘Animal pieces’’, and narrowed down our search with Iconclass to the subject ‘‘Classical Mythology and Ancient History’’. Finally, the merged view gives access to the collections through a merged thesaurus combining both original vocabularies into a single facet, based on the links found between them in the automatic mapping process. If we select the ARIA concept ‘‘Animal pieces’’, the view provides both ARIA concepts (such as ‘‘Birds’’) and Iconclass concepts (such as ‘‘29A:animals acting as human beings’’) for further refining our search.

Cultural heritage resources

195

Figure 7. Single view: using the ARIA thesaurus to browse the two collections

LR 57,3

196

Figure 8. Combined view: using ARIA and Iconclass to browse the two collections

Discussion and conclusion Existing alignment tools have been reported to perform poorly on real-life cases such as CH thesaurus alignment (Gendt et al., 2006). In fact, alignment is still an open research problem as no single technique is universally applicable, or will return satisfactory results. In practice, different techniques have to be carefully selected and combined, depending on the characteristics of the case at hand, such as the richness of the semantic structures of vocabularies, their lexical coverage and the existence of collections simultaneously described by several vocabularies. It should be noted, however, that a continuous improvement of techniques and tools can lead to significant improvements, as witnessed in the regular evaluation campaigns organised by the research community (Euzenat et al., 2006). The Semantic Web-inspired methods and tools described in this paper still require further experimentation in practical applications, and a greater availability of vocabularies. Nevertheless, current representation and alignment techniques can be employed to build demonstrators that showcase the integration of collections at the semantic level, leading the way from separate islands of collections and vocabularies to better connected networks of CH knowledge. One such demonstrator is described in the previous section of this paper; this faceted browser gives a unified access to two collections of illuminated manuscripts via any of its respective vocabularies. Other examples of web portals that illustrate the use of Semantic Web techniques in the CH domain can be seen on the websites of the MuseumFinland[15] and eCulture[16] projects. These projects, even if not focusing on semantic alignment, demonstrate the possible benefits of using Semantic Web technologies: the use of the SKOS representation format, the development of innovative interfaces to access Cultural Heritage collections, and the exploitation of automated reasoning techniques over RDF-based metadata. Other portals are being created with enhanced functionality and usability, as the synergy between CH and SW communities increases. One example is the ongoing eCulture project, which has been given the Semantic Web Challenge[17] award in 2006. In fact, the richness and high quality of CH data is very attractive to researchers of the Semantic Web community, as they have many tools but little real-life metadata to show their true potential. On the other hand, the CH domain (including digital libraries) could profit from techniques and tools developed by the SW community in creating a web of CH that delivers high quality content via easily accessible semantic search and navigation. Compared to current web searching techniques that are based on full text search (i.e. matching strings), Semantic search – matching meanings – represents a huge advance.

Notes 1. See www.theeuropeanlibrary.org (accessed 18 December 2007). 2. See www.geheugenvannederland.nl (accessed 18 December 2007). 3. The following is a simplified introduction to the Semantic Web. For further detail, the reader is encouraged to consult the Semantic Web Primer (Antoniou and Harmelen, 2004). 4. See www.w3.org/RDF (accessed 18 December 2007). 5. Figure 2 is an abstract representation of an RDF graph. Such a graph will be usually serialized in the form of an XML file, according to the RDF/XML syntax specified by the W3C. 6. Nodes in the graph are RDF resources; labelled edges represent assertions of a property between the linked elements. The rdf: namespace stands for http://www.w3.org/1999/ 02/22-rdf-syntax-ns#, rdfs: for http://www.w3.org/2000/01/rdf-schema#, myVoc1: for http://example.org/voc1#, and myVoc2: for http://example.org/voc2#. 7. SKOS is currently under scrutiny by the W3C Semantic Web Deployment Working Group and is planned to be published as a W3C Proposed Recommendation in 2008. See www.w3.org/2004/02/skos (accessed 18 December 2007). 8. In the following ‘‘skos:’’ stands for http://www.w3.org/2004/02/skos/core#. (accessed 18 December 2007). 9. See www.iconclass.nl (accessed 18 December 2007). 10. In this paper, alignment refers to the creation of semantic relationships (e.g. equivalence) between concepts coming from different KOSs in order to solve interoperability problems. This notion approximates what is referred to in the KOS community by vocabulary mapping, crosswalk or reconciliation, and in the Semantic Web community by ontology alignment, mapping or matching. 11. See www.gesis.org/en/research/information_technology/komohe.htm (accessed 18 December 2007). 12. See www.d-nb.de/wir/projekte/crisscross.htm (accessed 18 December 2007). 13. See http://macs.cenl.org (accessed 18 December 2007). 14. See www.opencyc.org (accessed 18 December 2007). 15. See www.museosuomi.fi (accessed 18 December 2007). 16. See http://e-culture.multimedian.nl (accessed 18 December 2007). 17. See http://challenge.semanticweb.org (accessed 18 December 2007). References Antoniou, G. and Harmelen, F. van (2004), Semantic Web Primer, MIT Press, Cambridge, MA. Balikova, M. (2005), ‘‘Multilingual subject access to catalogues of national libraries (MSAC), Czech Republic’s collaboration with Slovakia, Slovenia, Croatia, Macedonia, Lithuania and Latvia’’, in Proceedings of the 71st IFLA General Conference and Council ‘‘Libraries – A voyage of discovery’’, Oslo, Norway, 2005, available at: www.ifla.org/IV/ifla71/papers/044eBalikova.pdf (accessed 18 December 2007). Berners-Lee, T., Hendler, J. and Lassila, O. (2001), ‘‘The Semantic Web’’, Scientific American, Vol. 284 No. 5, pp. 34-43, available at: www.sciam.com/article.cfm?articleID=0004814410D2-1C70-84A9809EC588EF21 (accessed 18 December 2007). Brickley, D. and Guha, R.V. (Ed.), (2004), RDF Vocabulary Description Language 1.0: RDF Schema, W3C Recommendation, 10 February 2004, available at: www.w3.org/TR/rdfschema/ (accessed 18 December 2007).

Cultural heritage resources

197

LR 57,3

198

cDay, M., Koch, T. and Neuroth, H. (2005), ‘‘Searching and browsing multiple subject gateways in the Renardus service’’, in Proceedings of the Sixth International Conference on Social Science Methodology, Amsterdam. Doerr, M. (2001), ‘‘Semantic problems of thesaurus mapping’’, Journal of Digital Information, Vol. 1 No. 8, available at: http://jodi.tamu.edu/Articles/v01/i08/Doerr/ (accessed 18 December 2007). Euzenat, J. and Shvaiko, P. (2007), Ontology Matching, Springer, Berlin. Euzenat, J., Mochol, M., Shvaiko P., Stuckenschmidt, H., Svab O., Svatek, V., Hage, W. van and Yatskevich, M. (2006), ‘‘Results of the ontology alignment evaluation initiative 2006’’, in Proceedings of the First International Workshop on Ontology Matching, 5th International Semantic Web Conference (ISWC 2006), Athens, GA, USA, available at: www.dit.unitn.it/ ~p2p/OM-2006/7-oaei2006.pdf (accessed 18 December 2007). Giunchiglia, F., Shvaiko, P. and Yatskevich, M. (2005), ‘‘Semantic schema matching’’, in Proceedings of the 13th International Conference on Cooperative Information Systems (CoopIS 2005), Agia Napa, Cyprus, 2005. Harmelen, F. van (2005), ‘‘Ontology mapping: a way out of the medical tower of babel?’’, in Artificial Intelligence in Medicine: proceedings of the 10th Conference on Artificial Intelligence in Medicine (AIME 2005), Aberdeen, UK, 2005. Isaac, A., van der Meij, L. van der, Schlobach, S. and Wang, S. (2007) ‘‘An empirical study of instance-based ontology matching’’, in Proceeedings of the 6th International Semantic Web Conference (ISWC 2007), Busan, Corea, 2007. Jian, N., Hu, W., Cheng, G. and Qu, Y. (2005), ‘‘Falcon-AO: aligning ontologies with Falcon’’, in Proceedings of the K-CAP Workshop on Integrating Ontologies, Banff, Canada, 2005. Krause, J. (2003) ‘‘Standardization, heterogeneity and the quality of content indexing: a key conflict of digital libraries and its solution’’, in World library and information congress: 69th IFLA general conference and council, Berlin, 2003, available at: www.ifla.org/IV/ ifla69/papers/085e_trans-Krause.pdf (accessed 18 December 2007). Landry, P. (2004) ‘‘Multilingual subject access: the linking approach of macs’’, Cataloging and Classification Quarterly, Vol. 37 Nos. 3-4, pp. 177-91. Liang, A. and Sini, M. (2006) ‘‘Mapping AGROVOC and the Chinese agricultural thesaurus: definitions, tools, procedures’’, New Review in Hypermedia and Multimedia, Vol. 12 No. 1, pp. 51-62. Macgregor, G., McCulloch, E. and Nicholson, D. (2007) ‘‘Terminology server for improved resource discovery: analysis of model and functions’’, in Proceedings of the Second International Conference on Metadata and Semantics Research, Corfu, Greece, 2007, available at: http://eprints.cdlr.strath.ac.uk/3435/ (accessed 18 December 2007). McGuinness, D.L. and Harmelen F. van, (Ed.), (2004), OWL Web Ontology Language Overview, W3C Recommendation, 10 February 2004, available at: www.w3.org/TR/owl-features/ (18 December 2007). Miles, A. and Brickley, D. (2005), SKOS Core Guide, W3C Working Draft, 2 November 2005 (Work in progress), available at: www.w3.org/TR/swbp-skos-core-guide/ (accessed 18 December 2007). Miller, G. (1995), ‘‘Wordnet: a lexical database for English’’, Communications of the ACM, Vol. 38 No. 11, pp. 39-41. Shvaiko, P. and Euzenat, J. (2005), ‘‘Ontology matching’’, D-Lib Magazine, Vol. 11 No. 12, In Brief, available at: www.dlib.org/dlib/december05/12inbrief.html (accessed 18 December 2007). van Assem, M., Malaise, V., Miles, A. and Schreiber, G. (2005), ‘‘A method to convert thesauri to SKOS’’, in Proceedings of the Third European Semantic Web Conference, Budva, Montenegro, 2005.

van Gendt, M., Isaac, A., van der Meij, L. and Schlobach, S., (2006) ‘‘Semantic Web techniques for multiple views on heterogeneous collections: a case study’’, in Proceedings of the 10th European Conference on Research and Advanced Technology for Digital Libraries (ECDL 2006), Alicante, Spain, 2006. Zhang, X. (2006) ‘‘Concept integration of document databases using different indexing languages’’, Information Processing and Management, Vol. 42, pp. 121-35. Corresponding author Antoine Isaac can be contacted at: [email protected]

To purchase reprints of this article please e-mail: [email protected] Or visit our web site for further details: www.emeraldinsight.com/reprints

Cultural heritage resources

199

The current issue and full text archive of this journal is available at www.emeraldinsight.com/0024-2535.htm

LR 57,3

200

The AGROVOC Concept Server: rationale, goals and usage Margherita Sini, Boris Lauser, Gauri Salokhe, Johannes Keizer and Stephen Katz Food and Agriculture Organization of the United Nations (FAO), Rome, Italy

Received 19 October 2007 Reviewed 29 October 2007 Abstract Accepted 7 November 2007 Purpose – The main objective of the AGROVOC Concept Server (CS) is to create a collaborative

reference platform and a ‘‘one-stop’’ shop for a pool of commonly used concepts related to agriculture, containing terms, definitions and relationships between terms in multiple languages derived from various sources. This paper aims to address the issues. Design/methodology/approach – The CS offers a centralised facility where the agricultural information management community can build and share agricultural knowledge in a collaborative environment. Findings – The advantages of the CS are its extensibility and modularity that provide the possibility to extend the type of information that can be stored in this system based on user/community needs. Research limitations/implications – Further investigation still needs to be done on the modularisation of the CS (i.e. the creation of separated ontologies that can still be connected, in order to have domain-related ontologies and to allow for better performance of the CS). Practical implications – The CS serves as starting point for the development of specific domain ontologies where multilinguality and the localised representation of knowledge are essential issues. Furthermore, it will offer additional services in order to expose the knowledge to be consumed by other applications. Originality/value – The CS Workbench provides the AGROVOC partners with the possibility to directly and collaboratively edit the AGROVOC CS. It thus provides the opportunity for direct and open ‘‘many-to-many’’ communication links between communities, avoiding decentralised communication between partners and duplication of effort. For the international community, it may allow users to manage, re-use or extend agriculture-related knowledge for better interoperability and for improved services. Keywords Modelling, Information management, Information systems Paper type Conceptual paper

Introduction For over 30 years, the Food and Agriculture Organization of the United Nations (FAO) has facilitated a network of documentation centres from agricultural research and technology institutes and academic faculties. The aim of the network is to enhance the exchange of agricultural knowledge, especially between developing countries. The rise of the Internet in the 1990s has revolutionised the way people share and exchange knowledge. There is little doubt that the web provides a platform for global access to information; however, there are a number of important issues that need to be addressed for this potential to be fully exploited. The web was not initially envisioned as a tool for global access to information, and the underlying standards for information management are not entirely adequate. By the very nature of Internet architecture, information on similar subjects is scattered across many different servers around the world, based on different language, local differences and description. Library Review Vol. 57 No. 3, 2008 pp. 200-212 Emerald Group Publishing Limited 0024-2535 DOI 10.1108/00242530810865745

Copyright Food and Agriculture Organization of the United Nations. The views expressed in this publication are those of the authors and do not necessarily reflect the views of the Food and Agriculture Organization of the United Nations.

On the other hand, few tools are available to integrate related information from different sources and, as a result, it is often very difficult to find related information on the web. Such problems can only be solved if action is taken to establish appropriate norms, vocabularies, guidelines and standards to facilitate the integration of data from different sources, and to engage in effective data exchange. Through the adoption of international classification schemes, controlled vocabularies, open standards and common data models, these information management problems shall eventually be overcome. FAO, through the development of tools that exploit such standards, strives to provide an effective framework for ‘‘one-stop shopping’’, where people can search for agricultural information resources without having to explore many different individual websites. In the agricultural sector, there exist already many well-established and authoritative controlled vocabularies, such as FAOs multilingual thesaurus AGROVOC[1], the CAB Thesaurus[2] and the NAL Thesaurus[3] of the National Agricultural Library in the USA. These systems have been carried over from the traditional library world. However, for these semantic tools to be more effective on the Internet and especially in a multilingual context, there is a need to re-assess the traditional ‘‘thesaurus’’ approach and move towards more powerful models and technologies, such as the development of concept-based systems, known also as ‘‘ontologies’’ (Kang and Lee, 2001; Wielinga et al., 2001; Fisseha and Liang, 2003). In terms of the Semantic Web, ontology can be defined as a semantic system that describes concepts, the definitions of these concepts, and the specification of relationships amongst them. Ontology takes the traditional thesaurus approach one step further by structuring the terms more formally, by providing richer relationships between concepts and by adding other constraints on concepts that can be exploited by intelligent applications to reason on the concepts and to infer knowledge. Ontologies are an integral part of the Semantic Web, described by Tim Berners-Lee as ‘‘an extension of the current one, in which information is given well-defined meaning, better enabling computers and people to work in cooperation’’ (Berners-Lee et al., 2001). The AGROVOC Concept Server (CS) project has been conceptualised as a response to this approach and will be used to develop knowledge-based applications that better serve user needs, as already done by others (Clark et al., 2000). The AGROVOC CS is significantly different in that it functions as a resource to help structure and standardise agricultural terminology in multiple languages for use by any number of different users and systems around the world. In addition, the idea is to combine the CS, as a whole or extracted subject ontologies with application ontologies or metadata ontologies, in order to support the semantic infrastructure at FAO (Salokhe et al., 2004; Liang, 2006a). The idea of the CS Motivations The AGROVOC Thesaurus has been successfully used over decades in many systems inside and outside FAO. Its main use has been subject indexing of document like objects in digital libraries in the agricultural domain. In light of the aforementioned emerging technologies, FAO has been investigating the possibility of representing this knowledge in ways that are more suitable for the future web and that can help address the problems of information retrieval and integration (Soergel et al., 2004). The following points can be summarised as the primary motivations for undertaking the new developments:

AGROVOC Concept Server

201

LR 57,3

.

At present, the development and maintenance of the AGROVOC Thesaurus (translation, revision of relationships, suggestions of new terms) is centralised. All work from different partner organisations is directed to FAO for inclusion in the online version of AGROVOC. Given the time lag and cumbersome work carried out to update the online version, the idea was to create a new online version that can be maintained and updated directly by the partners. The distributed and collaborative maintenance would considerably reduce workflow overhead and duplication of effort.

.

The current structure of AGROVOC, which follows guidelines for multilingual thesauri, does not allow the users to add localised language information. AGROVOC is English centric meaning that all concepts start from an English version which is then translated into other languages. The idea of the new model is to provide users with the means to depict terms in their own local language, such as common names of plants, and create language specific relationships. This will create a much more powerful linguistic representation of the local knowledge.

.

One of the ultimate objectives is to provide better services to users, such as the possibility to retrieve information regardless of the language, spelling or term variants used for searching. Consequently, the introduction of the notion of a concept is required, in which one and the same concept is represented by a set of multiple terms in various languages that all identify the same conceptual idea. Information is hence associated with concepts rather than single terms.

.

The traditional and current AGROVOC thesaurus model, albeit useful in many contexts, is limited in its scope. The advances in data modelling, especially using the Ontology Web Language (OWL)[4], offer the possibility to create a concept based structure with explicit semantic relationships and other constraints that can be exploited for automatic inference of knowledge. Offering ways for modularisation, these new models allow the creation of sub-domain ontologies, using only some of the concepts. This allows the users to organise knowledge based on their own distinct application needs.

202

Compared to the traditional AGROVOC Thesaurus, the main characteristics of the CS are the following: .

It is a concept-based, modularised and extensible system.

.

It provides the possibility to realise term and language specific relationships, which offers for much more flexibility on the linguistic level.

.

It allows for the representation of more semantics in terms of concept and term relationships and other constraints and definitions provided by the OWL modelling language.

.

It accommodates distributed maintenance for improved workflow and better domain coverage.

Currently the CS model[5] is fully defined (Liang, 2006b) and available for reuse. An example of a domain specific ontology – implementing this model for the domain – is the crop wild relatives[6]. The full conversion of the AGROVOC Thesaurus based on this model is under development.

Goals and objectives With respect to the Semantic Web initiative, the CS strives to: .

provide a framework for sharing common terminology, concept definitions and relations within the agricultural community;

.

provide a powerful and extensible model that can be used to create other ontologies;

.

streamline efforts and enhance collaboration for the creation of knowledge management systems throughout the agricultural community;

.

increase the efficiency and consistency with which multilingual agricultural information objects and resources are described and associated together;

.

increase the functionality and relevance when accessing these resources.

Once fully completed and made operational, the CS will offer a contextually rich and modern framework for modelling, serving and managing agricultural terminology. When integrated with web-based search tools, it will greatly facilitate resource retrieval. It should provide access to document-like objects in a variety of languages and offer suggestions for other related resources that are potentially relevant to the topic of interest. The CS is foreseen to empower a variety of useful services such as automatic or semi-automatic translation services, information discovery and reasoning services, guided search services and concept disambiguation services. Such additional functionality will not only dramatically increase the scope of web-based search engines, but also revolutionise the way users interested in agricultural resources interact with the web. The overall goal is to improve worldwide access to agricultural information and, as such, the CS, in combination with FAO’s other standardisation activities, plays a strategic role in its effort to fight hunger with information. Scope The CS operates in the agricultural domain (covering agriculture, fisheries, forestry, nutrition etc.) and provides the possibilities to mix-and-match the concepts and their relationships into specific sub-domain ontologies. The maintenance tool of the CS, will assure regular production of AGROVOC as a traditional thesaurus for systems that wish to continue using AGROVOC as a thesaurus. The CS will not necessarily integrate or manage ontologies which are created outside the scope of the CS. However, it may provide links to download these ontologies. The CS will provide web services that can be consumed by application developers to be used to enhance their systems. Comparison with other approaches Other models and approaches have already stated the necessity to identify the concept rather than the term as done in traditional thesauri. The ISO Standard 5964[7] (Multilingual Thesauri) clearly states the issues related to term translation. Recently, the British Standard (BS) 8723 has been circulated amongst small expert groups in the areas of SKOS and XML. A thesaurus data model for the BS is already available in UML[8]. Simple knowledge organisation systems (SKOSs)[9] and other similar approaches also take into consideration the notion of the ‘‘enhanced thesaurus’’, so the introduction of concepts vs terms (and more refined relationships) is nothing new to the approach presented here.

AGROVOC Concept Server

203

LR 57,3

204

Our model has evolved in a long process of research and collaboration using feedback from the agricultural community, so it represents a direct response to the requirements gathered over years from this community and domain. The CS and its maintenance model have been developed with the goal of providing a model that is powerful enough to be extended beyond what thesauri and thesaurus applications can offer. For instance, these approaches introduce the notion of concepts, multilinguality and all the mentioned linguistic advances, but also further constraints (e.g. the possibility for instantiation of concepts and other features that are permitted in OWL). So the strength is in the extensibility. It will always be possible to simplify the model and interface with others such as SKOS, the new British Standard or others. Moving towards the CS The CS is an element of the AOS initiative and the core server of knowledge for the new services that FAO, together with partners, would like to offer to its users. The process to move towards the CS has been a gradual one. Over the past few years, FAO and its partners have been carrying out activities that support the transformation of the AGROVOC Thesaurus into the CS. In particular, the following activities can be pointed out as being crucial. The idea of the CS first arose in 2001 with the main goal of integrating all the FAO terminological resources. This idea, however, was not very successful because of their disparate structures, scopes and owners. However, the need to restructure AGROVOC as ontology was carried forward. The following figure (Figure 1) depicts the initial vision of the Agricultural Ontology Server (AOS). Back then the AOS was envisioned to be what today is represented by the CS. In 2002, the first prototype domain ontologies were developed (fisheries, food safety), experimenting with tools to automatically extract concepts and domain

Figure 1. AOS vision (2001)

specific knowledge of the AGROVOC Thesaurus and using domain experts to validate and develop it further. Between 2002 and early 2005, a study was conducted to identify the form and technical infrastructure of the CS. The main evaluation was on the possibility of storing an ontology in a relational database (Soergel, 2004). In 2005, research was carried out to use the OWL for representing the CS structure and data. During this period, OWL was gaining widespread interest from a range of disciplines and domains, including medicine, defence, agriculture, biology and library sciences. More sophisticated and better performing technologies were continually being developed on top of the OWL ontologies. The research provided the following reasons to move from a relational database model to OWL: (1) OWL allows for easy integration of other RDF-based data sources at the storage level. (2) OWL allows for straightforward data processing and visualisation. (3) The OWL model is reusable and interoperable with any RDF triple-store. (4) OWL is web-enabled which would make data transfer and reuse much more likely. It is also a W3C recommendation. (5) OWL tools may be reused. For example, reuse and modification of open source tools that use OWL (e.g. Prote´ge´, SWOOP, etc.). These tools already implement many of the functionalities needed for the CS maintenance system. (6) OWL is an evolving and recognised standard which will assure for maximum interoperability and future support from a wider community. Subsequently, several efforts were made to represent AGROVOC in SKOS and OWL. In 2006, the AGROVOC CS OWL model was proposed. FAO’s research has identified three levels of information, which need to be represented in the CS (Liang et al., 2006b): .

Concept level: a specific notion, concrete or abstract. Every concept will be identified by one or more definitions.

.

Term level: any specific sense associated to a concept (every language or synonym is represented as a separate term).

.

String level: all word variances representing the same term (here we may include singular and plural forms, spelling variances, abbreviations, etc.).

The power of the semantics comes into place with the possibility to explicitly link concepts with relationships, such as ‘‘has pest’’ or ‘‘ingredient of’’. Such relationships enable the user to learn more about the particular concept and explore the domain around it by following the relationships. Terms may be connected to each other by the identification of exact translations, synonyms or other linguistic information, which may be used to build semi-automatic translation systems. Strings may be differentiated and connected by specifying for example uses of a form (e.g. long forms vs a short one). This new structure allows us to refine and to enhance the information represented originally within a traditional thesaurus. A revision and refinement of the AGROVOC Thesaurus terms and relationships was carried out, during which many terms were added or revised in order to provide users a better pool of data. All generic thesaurus relationships correspond to multiple meanings (e.g. BT/NT does not necessarily correspond to ‘‘subclass of’’, but to ‘‘part of’’ or any other specific relationship). Therefore, a specific tool was developed to help

AGROVOC Concept Server

205

LR 57,3

206

experts in revising and refining the thesaurus relationships into more meaningful and more specific semantic relationships. To accommodate linguistic issues like the identification of acronyms, synonyms and other term variances, or specific terms such as all chemical-related terms, a revision of the term relationships (mainly between descriptors and non-descriptors) and scope[10] was necessary. This activity is only partially completed and currently ongoing. Furthermore, better organisation of scientific and common names for taxonomic entities is necessary. The current AGROVOC Thesaurus does not clearly distinguish between taxonomic terms and common names; they may be related to each other with the ‘‘Related Term’’ relationship, but may not indicate that they really describe the same concept. Currently, the development of a tool for the collaborative maintenance of the data pool is ongoing (CS Workbench). A preliminary version is available for demonstration and is undergoing testing[11]. In parallel, several projects have been undertaken to extract specific sub-domain ontologies from AGROVOC for use in specific information systems or applications, both inside and outside FAO. These include: .

Food safety ontology (Lauser et al., 2002; Volz et al., 2003)

.

Food, Nutrition and Agriculture portal (Sini et al., 2007b)

.

Fishery ontology (Gangemi et al., 2002; Gangemi et al., 2004)

.

AGROVOC Topic Map (Kawtrakul et al., 2007)

.

Fertilizers ontology (Sanchez-Alonso and Sicilia, 2007)

.

Crop Wild Relative ontology (Morten, 2007)

The CS and its role in AOS today Over the years, the initial idea of the AOS developed into something much larger. The Agricultural Ontology Service, of which the CS is now an integral part, also includes domain ontologies, registries of mappings, URN services to name but few. Figure 2 below depicts the overall architecture of the ‘‘new’’ Agricultural Ontology Service. It hosts a wide variety of elements and services which are necessary to realise interoperability in the agricultural domain and which will be made available to users in the international community for better harmonisation of data and better tools development. In Figure 2: (1) The CS as described above represents the core of the AOS. (2) A knowledge organisation system (KOS) registry will be maintained in order to register trusted and well-developed KOS within the agricultural community (whether based on the CS model or created using other models). (3) A registry of mappings will be maintained through which mappings between featured KOS will be made available for use to be incorporated into other systems for disambiguation, translation and other purposes. (4) Other ontologies and KOS developed under the umbrella of the AOS (Fisheries Ontologies, Nutrition, Crop Wild Relatives, etc.) will be made available through AOS. Additional, AOS Services will include (but are not limited to): .

a URN registry;

AGROVOC Concept Server

207

Figure 2. Agricultural Ontology Service vision (2007) and the role of CS within

.

a relationship registry for approved relationships within the AOS and agricultural community;

.

multi-KOS search for keyword and semantic searches across several KOS, exploiting mappings and other features of the model;

.

web services to access hierarchies, concept details, linguistic information, relationships, etc. of the CS or other featured KOS.

Community and roles The AGROVOC Thesaurus is currently used and maintained by different partner organisations around the world. However, there is lack of synchronisation in content creation and updating, and there are only generic and poor common guidelines for standard actions or good practice (Sini et al., 2007a). Incorporating local knowledge and other languages is therefore cumbersome and slow.

LR 57,3

208

Figure 3 shows the current workflow for updating AGROVOC and managing translations and updates in a centralised environment. The CS Workbench is a collaborative tool for ontology management. It can be used by different users who may have different privileges and roles. The different roles that have been identified and are currently implemented are as follows: (1) Not logged-in users: These users can browse or search the system and submit suggestions for terms in any language, concepts and relationships. They can also provide other comments and feedback on the system in general. (2) Term editors (terminologist/thesaurus editor): Term editors are content experts in specific domains related to agriculture. They have full permissions on the term level in their assigned languages but have no rights for concepts modifications; however, they can still make suggestions for concept modifications. (3) Ontology editors (more experienced terminologist/thesaurus editor): Ontology editors are experienced in ontology modelling and are familiar with the concept-term-string level. They have full permissions to manage concepts and terms in their assigned languages. (4) Validators: Validators are experts that can check and validate the work done by the editors (terms and ontology editors) depending on their assigned languages. They should be experienced both in agricultural content and ontology modelling practices. (5) Publishers (ontology editors): Publishers have full permission on terms and concepts in all languages. They generally confirm the work of the validators, but can directly validate ontology editors and term editors also. They are the final instance in the validation workflow. (6) Administrators: Administrators have full access to all system functionalities and all languages. They will manage users, groups and permissions, provide statistics and perform other duties.

Figure 3. Current workflows for translations and maintenance of AGROVOC

The current AGROVOC partners will be contacted to serve the different roles as individuals or as an organisation. This new collaborative infrastructure will significantly change the current workflow drawn in Figure 3. The new infrastructure proposes a system, in which all actors will interact collaboratively and concurrently (Figure 4). The collaboration in this case is much more effective because: .

AGROVOC editors all over the world can have direct access to the maintenance tool;

.

changes are immediate and there is no need to wait for FAO actions (apart from the validation phase which will be carried out by FAO);

.

all users can immediately see and benefit from users’ contributions;

.

the cycle of adding data to AGROVOC and reusing it in the respective systems is more immediate; after data is inserted in the system, and eventually validated, it is immediately available for remote access through web services or for download in various formats.

AGROVOC Concept Server

209

Owing to the collaborative nature of the CS and the number of people that will eventually interact via the new infrastructure, well defined and managed workflows are necessary to avoid confusion, data inconsistency and to ensure quality control. A validation workflow and user rights management system is currently being implemented in the CS Workbench. For example, only Administrators will have the right to edit, add or delete the list of relationships that can be used to link concepts and terms. Intellectual property rights (IPR) and data custody issues The IPR, in particular copyright, of material such as terminological data, glossaries, images, and so forth, shall remain with the originating party, who will be indicated as the source partner if the information is reproduced or disseminated through the CS or elsewhere. Copyright of the information, as well as rights to any other intellectual

Figure 4. Proposed infrastructure for collaborative CS management and use

LR 57,3

210

property, developed in collaboration within the network shall be jointly vested to all parties involved. Each party shall have full rights to exploit such jointly owned works after informing the other parties, without the need of approval from the other. All partners to the network shall have free access to all information developed. As the IPR issues are extremely important to ensure that work is rightfully recognised, this area will be further discussed by the partners in the next few months of development. Conclusion and future work The CS is conceived as a pool of semantically related concepts. All concepts are represented with multiple terms and definitions in many languages. Multilinguality is addressed in a very complete but extensible conceptual model. The model is capable of arranging the complex multilinguality as managed by FAO (even accounting for nonLatin languages) and terminology information. Terms have relationships that allow the identification of exact translations of terms, synonym relationships or word variances. A major advantage of the CS is its extensibility: The amount and type of information that can be stored in this system can be extended anytime based on needs. For example, at the term level, we could add information about the term etymology, which will allow AOS to provide additional services to specific users. Other users could take the CS or parts of it as a basis and extend the model to make it ‘‘OWL Full’’. In this way, users would be able to treat concepts as instances and express even more sophisticated knowledge according to their needs. The CS is one of the key elements of the AOS initiative, which aims to provide better services to users of the agricultural domain exploiting new semantic technologies. The full AOS initiative will provide additional tools and services for the exploitation of the data contained in the CS (e.g. tools for semi-automatic generation of taxonomies or domain ontologies to be used as a basis for new FAO websites, web services for accessing CS data, tools for semi-automatic translation, for semi-automatic indexing, etc.). The full AGROVOC Thesaurus has been converted to a unique ontology, which can be further modularised by top concepts, categories or classification schemes. However, further investigation is required with respect to the modularisation of the CS (i.e. the creation of separated ontologies that can still be connected, in order to have domainrelated ontologies and to allow for better performance of the CS). Notes 1. AGROVOC Thesaurus: www.fao.org/aims/ (accessed 18 December 2007). 2. CAB Thesaurus: www.cabi.org/DatabaseSearchTools.asp?PID ¼ 277 (accessed 18 December 2007). 3. NAL Thesaurus: http://agclass.nal.usda.gov/dne/search.shtml (accessed 18 December 2007). 4. Ontology Web Language www.w3.org/2007/OWL/wiki/OWL_Working_Group (accessed 18 December 2007). 5. Concept Server Model: www.fao.org/aims/aos/aos.owl (accessed 18 December 2007). 6. Crop Wild Relatives Ontology: www.fao.org/aims/aos/cwr_DL.owl (accessed 18 December 2007). 7. ISO 5964:1985 Documentation – Guidelines for the establishment and development of multilingual thesauri www.collectionscanada.ca/iso/tc46sc9/standard/5964e.htm (accessed 18 December 2007). 8. http://isegserv.itd.rl.ac.uk/blogs/alistair/archives/40 (accessed 18 December 2007).

9. Simple knowledge organisation systems (SKOS): www.w3.org/2004/02/skos/ (accessed 18 December 2007). 10. Scope is specific information of the AGROVOC thesaurus identifying if a term belong to a specific sub-vocabulary such as geographical terms, taxonomical terms or chemicals. 11. http://vivaldi.cpe.ku.ac.th:8085/agrovoc/ (accessed 18 December 2007). References Berners-Lee, T., Hendler, J.A. and Lassila, O. (2001), ’’The Semantic Web’’, Scientific American, May 2001, available at: www.sciam.com/article.cfm?articleID ¼ 00048144-10D2-1C7084A9809EC588EF21 (accessed 18 December 2007). Clark, P., Thompson, J., Holmback, H. and Duncan, L. (2000), ‘‘Exploiting a thesaurus based semantic net for knowledge-based search’’, in Proceedings of IAAI-2000, available at: http://citeseer.ist.psu.edu/309406.html (accessed 18 December 2007). Fisseha, F. and Liang, A.C. (2003), ‘‘Reengineering AGROVOC to ontologies: steps towards better semantic structure, NKOS Workshop, 31 May 2003, Rice University, Houston, available at: ftp://ftp.fao.org/gi/gil/gilws/aims/publications/presentations/2003_7.pdf (accessed 18 December 2007). Gangemi, A., Fisseha, F., Keizer J., Liang, A., Pettman, I., Sini, M. and Taconet, M. (2004), ‘‘A core ontology of fishery and its use in the fishery ontology service project’’, in Proceedings of the EKAW*04 Workshop on Core Ontologies in Ontology Engineering, 8 October, Northamptonshire, UK, available at: http://ftp.informatik.rwth-aachen.de/Publications/ CEUR-WS/Vol-118/paper4.pdf (accessed 18 December 2007). Gangemi, A., Fisseha, F., Pettman, I. and Keizer, J. (2002), ‘‘Building an integrated formal ontology for semantic interoperability in the fishery domain’’, Agricultural Information and Knowledge Management Papers, available at: ftp://ftp.fao.org/docrep/fao/008/af242e/ af242e00.pdf (accessed 18 December 2007). Hulden, M. (2007), ‘‘A practical approach on creating a restricted ontology for crop wild relatives’’, Agricultural Information and Knowledge Management Papers, available at: ftp:// ftp.fao.org/docrep/fao/010/ah862e/ah862e.pdf (accessed 18 December 2007). Kang, S.J. and Lee, J.H. (2001), ‘‘Semi-automatic practical ontology construction by using a thesaurus, computational dictionaries, and large corpora’’, in Proceedings of the Workshop on Human Language Technology and Knowledge Management (Annual meeting of the ACL), available at: http://portal.acm.org/citation.cfm?id ¼ 1118226 (accessed 18 December 2007). Kawtrakul, A., Yingsaeree, C. and Andres, F. (2007), ‘‘A framework of nlp based information tracking and related knowledge organizing with topic maps’’, in Natural Language Processing and Information Systems (Lecture Notes in Computer Science), SpringerLink, Berlin/Heidelberg, pp. 272-83. Lauser, B., Wildemann, T., Poulos, A., Fisseha, F., Keizer, J. and Katz, S. (2002), ‘‘A comprehensive framework for building multilingual domain ontologies: creating a prototype biosecurity’’, in Proceedings of the International Conference DC-2002: Metadata for e-Communities: ‘‘Supporting Diversity and Convergence’’, 13-17 October, Biblioteca Nazionale Centrale Firenze, available at: www.bncf.net/dc2002/program/ft/paper13.pdf (accessed 18 December 2007). Liang, A., Salokhe, G., Sini, M. and Keizer, J. (2006a), ‘‘Towards an infrastructure for semantic applications: methodologies for semantic integration of heterogeneous resources’’, Cataloging and Classification Quarterly, Vol. 43 Nos. 3-4, pp. 161-89, available at: ftp:// ftp.fao.org/docrep/fao/009/ag869e/ag869e00.pdf (accessed 18 December 2007). Liang, A., Lauser, B., Sini, M., Keizer, J. and Katz, S. (2006b), ‘‘From AGROVOC to the agricultural ontology service/concept server: An OWL model for managing ontologies in the

AGROVOC Concept Server

211

LR 57,3

212

agricultural domain’’, in Proceedings of OWL: Experiences and Directions Workshop, 1011 November, Atlanta, available at: http://owl-workshop.man.ac.uk/acceptedPosition/ submission_31.pdf (accessed 18 December 2007). Salokhe, G., Pastore, A., Richards, B., Weatherley, S., Aubert, A., Keizer, J., Nadeau, A., Katz, S., Rudgard, S. and Mangstl, A. (2004), ‘‘FAO’s role in information management and dissemination – challenges, innovation, success, lessons learned’’, Quarterly Bulletin of the International Association of Agricultural Information Specialists (IAALD), Vol. 49 Nos. 3-4, pp. 73-83, available at: www.fao.org/docrep/008/af238e/af238e00.htm (accessed 18 December 2007). Sanchez-Alonso, S. and Sicilia, M.A. (2007), ‘‘Using an AGROVOC-based ontology for the description of learning resources on organic agriculture’’, in Proceedings of the Second International Conference on Metadata and Semantics Research (MTSR’07), Corfu, Greece, 11-12 October, available at: www.cc.uah.es/ssalonso/papers/AGOVOC_mtsr.pdf (accessed 18 December 2007). Sini, M., Johannsen, G. and Salokhe G. (2007a), Basic Guidelines for Managing AGROVOC, FAO, Rome, available at: ftp://ftp.fao.org/docrep/fao/010/ai144e/ai144e00.pdf (accessed 18 December 2007). Sini, M., Salokhe,G., Pardy, C., Albert, J., Keizer, J. and Katz, S. (2007b), ‘‘Ontology-based navigation of bibliographic metadata: example of the Food, Nutrition and Agriculture Journal’’, in Proceedings of the International Conference on the Semantic Web and Digital Libraries (ICSD-2007), 21-23 February, DRTC, Bangalore, available at: ftp://ftp.fao.org/ docrep/fao/009/ah765e/ah765e00.pdf (accessed 18 December 2007). Soergel, D., Lauser, B., Liang, A., Fisseha, F., Keizer, J., Katz, S. (2004), ‘‘Reengineering thesauri for new applications: the AGROVOC example’’, Journal of Digital Information, Vol. 4 No. 4, available at: http://jodi.tamu.edu/Articles/v04/i04/Soergel/ (accessed 18 December 2007). Volz, R., Studer, R., Maedche, A. and Lauser, B. (2003), ‘‘Pruning-based identification of domain ontologies’’, in Proceedings of I-KNOW ‘03, 2-4 July, Graz, Austria, available at: http:// i-know.know-center.tugraz.at/previous/i-know03/papers/kc/volz.pdf (accessed 18 December 2007). Wielinga, B.J., Schreiber, A. Th., Wielemaker, J. and Sandberg, J.A.C. (2001), ‘‘From thesaurus to ontology’’, in Proceedings of the First international conference on Knowledge capture, ACM Press, Victoria, British Columbia, pp. 194-201, available at: www.cs.vu.nl/~guus/papers/ Wielinga01a.pdf (accessed 18 December 2007). Corresponding author Margherita Sini can be contacted at: [email protected]

To purchase reprints of this article please e-mail: [email protected] Or visit our web site for further details: www.emeraldinsight.com/reprints

The current issue and full text archive of this journal is available at www.emeraldinsight.com/0024-2535.htm

Reducing semantic complexity in distributed digital libraries Treatment of term vagueness and document re-ranking Philipp Mayr, Peter Mutschke and Vivien Petras GESIS-IZ Social Science Information Centre, Bonn, Germany Abstract

Reducing semantic complexity 213 Received 19 October 2007 Reviewed 9 November 2007 Accepted 13 November 2007

Purpose – The general science portal ‘‘vascoda’’ merges structured, high-quality information collections from more than 40 providers on the basis of search engine technology (FAST) and a concept which treats semantic heterogeneity between different controlled vocabularies. First experiences with the portal show some weaknesses of this approach which come out in most metadata-driven Digital Libraries (DLs) or subject specific portals. The purpose of the paper is to propose models to reduce the semantic complexity in heterogeneous DLs. The aim is to introduce value-added services (treatment of term vagueness and document re-ranking) that gain a certain quality in DLs if they are combined with heterogeneity components established in the project ‘‘Competence Center Modeling and Treatment of Semantic Heterogeneity’’. Design/methodology/approach – Two methods, which are derived from scientometrics and network analysis, will be implemented with the objective to re-rank result sets by the following structural properties: the ranking of the results by core journals (so-called Bradfordizing) and ranking by centrality of authors in co-authorship networks. Findings – The methods, which will be implemented, focus on the query and on the result side of a search and are designed to positively influence each other. Conceptually, they will improve the search quality and guarantee that the most relevant documents in result sets will be ranked higher. Originality/value – The central impact of the paper focuses on the integration of three structural value-adding methods, which aim at reducing the semantic complexity represented in distributed DLs at several stages in the information retrieval process: query construction, search and ranking and re-ranking. Keywords Digital libraries, Worldwide web, Information management Paper type Research paper

Introduction In the area of scientific and academic information systems, a whole array of bibliographic databases, disciplinary Internet portals, institutional repositories or archival and other media type collections are increasingly accumulated and embedded in all-encompassing information systems. Such collections are necessary in order to meet user expectations that demand one-stop ‘‘information fulfillment’’. Examples are Elsevier’s Scirus portal[1], the Online Computer Library Center WorldCat union catalog[2] or Tuft University’s Perseus project[3]. In Germany, an ambitious project for one-stop academic search is the vascoda portal[4], a joint project between the BMBF (Federal Ministry for Education and Research) and the DFG (German Research Foundation). Vascoda provides a federated search interface for a multitude of disciplinary and interdisciplinary databases (e.g. full-text article databases, indexing and abstracting services, library catalogs) and internet resource collections. The vascoda portal contains many information collections that are meticulously developed and structured. They have sophisticated subject metadata schemes (subject headings, thesauri or classifications) to describe and organise the content of the

Library Review Vol. 57 No. 3, 2008 pp. 213-224 # Emerald Group Publishing Limited 0024-2535 DOI 10.1108/00242530810865484

LR 57,3

214

Figure 1. Two step methodology of vagueness treatment

documents on an individual collection level. The general search interface, however, only provides a free-text search over all metadata fields without regard for the precise subject access tools that were originally intended for these information collections. If large-scale contemporary information organisation efforts like the Semantic Web[5] (see also Krause, 2006, 2007, 2008) strive to provide more structure and semantic resolution with respect to information content, how is it possible that advanced interfaces for digital libraries (DLs) scale back on exactly the same issue?. Search – both in full-text collections like the Internet or more heavily structured and less diverse collections like institutional repositories, indexing databases or library catalogs as described above – only works as well as the matching between the language in queries and the language in the searched documents. If the words in the query are different from the words in a relevant document, this document will not be found. The problem of matching query terms to document terms is a result of the ambiguity or vagueness of language (Blair, 1990, 2003). Because of the sheer size and variation of large full-text databases, this problem is not as noticeable because any query (even if they contain spelling mistakes or nonsense statements) will find documents. The problem is aggravated in collections of more restricted volume or text (i.e. repositories that contain only formal metadata, some subject description and just a link to the full-text). The issue becomes even more critical when several collections with different metadata schemes are searched at the same time – which is the case in the distributed search scenario. In this scenario, not only is the matching between query and document terms affected by language ambiguity, but also the matching between different subject-describing metadata schemes. In Figure 1, we speak of vagueness 1 and vagueness 2/3 (V1 and V2/3) to denote the different areas where language ambiguity can occur. For successful retrieval in any DL, both levels of vagueness have to be addressed (compare Hellweg et al., 2001). Furthermore, the result sets of transformed or expanded queries in distributed collections are often very large and tests show that the conventional web-based ranking methods are not appropriate for the heterogeneous metadata records. Therefore, two methods, which are derived from scientometrics and network analysis, will be implemented with the objective to re-rank result sets: (a) the ranking of the results by core journals (so-called Bradfordizing) and (b) ranking by centrality of authors in co-authorship networks.

This paper is a description of an attempt to harness the semantic knowledge in controlled vocabularies for several stages in the information retrieval process: query construction, search and ranking and re-ranking. We briefly describe the GESIS project ‘‘Competence Center Modeling and Treatment of Semantic Heterogeneity’’ with the goal of creating a semantic network of terms in different controlled vocabularies (terminology mapping) in order to facilitate a seamless search across different subjectbased knowledge organisation systems. At the conclusion of this project, we will discuss modules that are being devised to leverage these mappings for an improved user search experience. Results from a major terminology mapping effort Semantic integration seeks to connect different information systems through their subject metadata frameworks; insuring that distributed searches over several information systems can still use the advanced subject access tools provided with the individual databases. Through the mapping of different subject terminologies, a ‘‘semantic agreement’’ for the overall collection to be searched is achieved. Terminology mapping – the mapping of words and phrases of one controlled vocabulary to the words and phrases of another – creates a semantic network between the information systems carrying the advantages of controlled subject metadata schemes into the distributed DL world. In 2004, the German Federal Ministry for Education and Research funded a major terminology mapping initiative at the GESIS Social Science Information Centre in Bonn (GESIS-IZ) ‘‘Competence Center Modeling and Treatment of Semantic Heterogeneity’’[6], which concluded this year (see Mayr/Walter, 2007a, b). The task of this terminology mapping initiative was to organise, create and manage ‘‘crossconcordances’’ between major controlled vocabularies (thesauri, classification systems, subject heading lists), centred around the social sciences but quickly extending to other subject areas (e.g. political science, economics, medicine or subject-specific parts of universal vocabularies). Cross-concordances are intellectually (manually) created crosswalks that determine equivalence, hierarchy and association relations between terms from two controlled vocabularies. Most vocabularies in the project have been related bilaterally; that is, there is a cross-concordance relating terms from vocabulary A to vocabulary B as well as a cross-concordance relating terms from vocabulary B to vocabulary A (note: bilateral relations are not necessarily symmetrical). Other definitions and examples of crosswalks between controlled vocabularies exist in an international context (see overview in Zeng/Chan, 2004; Vizine-Goetz et al., 2004; Liang and Sini, 2006). In November 2007, 25 controlled vocabularies from 11 disciplines were connected with vocabulary sizes ranging from 1,000 to 17,000 terms per vocabulary. To date, more than 513,000 relations in 64 crosswalks have been generated. An overview of the preliminary project results presented at the NKOS/ECDL workshop 2007 can be found in[7]. A database including all mapped controlled terms and cross-concordance relations was built and a ‘‘heterogeneity service’’ developed. The heterogeneity service is a web service, which makes the cross-concordances available for other applications (see Figure 2). Many cross-concordances are already implemented and utilised for the German Social Science Information Portal sowiport[8], which searches bibliographical and other information resources (including 13 databases with 10 different vocabularies and about 2.5 million references).

Reducing semantic complexity 215

LR 57,3

216

Figure 2. Heterogeneity service (HTS)

Semantic mappings could support distributed searching in several ways. First and foremost, they should enable seamless searching in databases with different subject metadata systems. Additionally, they can serve as tools for vocabulary expansion since they present a vocabulary network of equivalent, broader, narrower and related term relationships. Thirdly, this vocabulary network of semantic mappings can also be used for query expansion and reformulation. The following section introduces the concept of a search term recommender (STR). This tool is an aid for query reformulation and reconstruction which has been adapted for a web-based information portal from human search intermediaries (e.g. reference librarians). Search term recommender Semantic mappings can reduce the problem of language ambiguity at the vagueness 2/3 layer described in Figure 1 (between different information systems). However, the vagueness at the user-information system interface remains unaddressed. To reduce language ambiguity at the vagueness 1 layer (between query terms and document terms), another instrument is necessary to map terms at this interface. The goal of this mapping would be to ‘‘translate’’ the query terms of a user to the document terms of the database (or vice versa) in order to produce a match at search time. Since we are mostly concerned with information systems that contain sparse text enriched with subject describing controlled vocabularies, we propose a STR system which will propose terms from the controlled vocabularies, given a specified query. The basic parameters of a search term suggestion system are the controlled vocabulary terms that are used for document representation and the natural language keywords that are input by the searcher. The advantage of suggesting controlled vocabulary terms as search terms is that these terms have been systematically assigned to the documents, so that there is a high probability of relevant and precise

retrieval results if these terms are used instead of whatever natural language keywords the searcher happens to think of. A second advantage in suggesting controlled vocabulary terms is their application in the semantic network of the cross-concordances. That is, if controlled vocabulary terms are used in searching, the cross-concordances, which map these terms between different databases, can be successfully applied for distributed retrieval. In addition, this kind of vocabulary help will hopefully improve the search experience for the user in general. Suggesting terms reduces the searcher’s need to think of other relevant search terms that might describe his or her information need. It effectively eases the cognitive load on the searcher since it is much easier for a person to pick appropriate search terms from a list than to come up with search terms by themselves. It also helps to alleviate ‘‘anchoring bias’’ (Blair, 2002), which is an effect that makes it harder to substantially deviate from one’s original thought-of search terms and to consider different search terms or strategies. Another consequence of term suggestion is the presentation of new or different technical expressions for a concept. This again could lead to changes in a search strategy or topic, which might help in reaching the user search goal. Term suggestions from several fields of research and/or information resources could also provide an overview over different areas of discussion, which deal with particular concepts (perhaps assuming different meanings or directions of thought). The result would be a different domain perspective on certain concepts, an effect which can also be achieved by displaying the semantic mappings of the cross-concordances themselves. A STR is created by building a dictionary of associations between two vocabularies: (1) natural language terms and phrases from the documents in the information collection (e.g. titles, abstracts, authors) and, (2) the controlled vocabulary (thesaurus terms, subject headings, classification numbers, etc.) used for document representation. In one implementation, a likelihood ratio statistic is used to measure the association between the natural language terms from the collection and the controlled vocabulary terms to predict which of the controlled vocabulary terms best mirror the topic represented by the searcher’s search terms (Plaunt and Norgard, 1998; Gey et al., 1999). However, other methods of associating natural language terms and controlled vocabulary terms are possible (Larson, 1991, 1992). In an information system with several information resources (i.e. databases) and several controlled vocabularies, a search term recommendation tool has to determine which terms from which vocabularies to suggest to the user and how to tie the term suggestions for query construction into the semantic network of cross-concordances. Several approaches seem possible: a pivot controlled vocabulary, from which terms are suggested and mappings approached; a general suggestion pattern, which clusters similar concepts from several vocabularies; or a domain-specific approach, whereby terms and vocabularies are chosen according to the subject of interest for the searcher. The result sets of transformed or expanded queries in distributed collections are often very large and tests show that the conventional web-based ranking methods are not appropriate for presenting heterogeneous metadata records as suitable result sets to the user. In the following section, we propose re-ranking methods (implemented as post-search modules), which are based on structures and regularities in scientometrics and network analysis.

Reducing semantic complexity 217

LR 57,3

218

Re-ranking Compared to traditional text-oriented sorting mechanisms, our scientometric and network analysis re-ranking methods offer a completely new view on results sets, which have not been implemented in heterogeneous and larger database scenarios to date. The usage of these modules should be an alternative ranking opportunity with the objective to enhance and improve the search process in general. In addition, we expect an improvement in document relevance for the top-listed documents. Bradford Law of Scattering and Bradfordizing Bradford Law of Scattering and Bradfordizing have their roots in scientometrics and are often applied in bibliometric analyses of databases and collections as a tool for systematic collection management in library and information science. Fundamentally, Bradford Law states that literature on any scientific field or subject-specific topic scatters in a typical way. A core or nucleus with the highest concentration of papers (a few core journals) on a topic is followed by zones with loose concentrations of paper frequency, which is described by Bradford: . . . if scientific journals are arranged in order of decreasing productivity of articles on a given subject, they may be divided into a nucleus of periodicals more particularly devoted to the subject and several groups or zones containing the same number of articles as the nucleus, when the numbers of periodicals in the nucleus and succeeding zones will be as 1:n:n2 . . . (Bradford, 1948)

Bradford Law as a general law in informetrics can be applied to all scientific disciplines, and especially in a multi-database scenario and in combination with the aforementioned semantic treatment of heterogeneity. Bradfordizing (White, 1981) is an information science application of the Bradford Law of Scattering which sorts/re-ranks a result set according to the identified core journals for a query. The journals for a search are ranked by the frequency of their listing in the result set (number of articles for a journal title). If a search result is bradfordized, articles of core journals are ranked ahead of the journals which contain an average number or only few articles on a topic. This method is interesting in the context of our re-ranking task because it is a robust way of sorting the central publication sources for any query to the top positions of a result set. Bradfordizing has the following values-added: (1) An alternative view on results sets which are ordered by core journals (the user is provided with documents of core journals first); (2) An alternative view on publication sources in an information space which are intuitively closer at the research process than statistical methods (e.g. best match) or traditional methods (e.g. exact match); (3) Possibly a higher topical relevance of re-ranked documents. Additionally, re-ranking via bradfordized lists offer an opportunity to switch between term-based search and the alternative search mode browsing. Bates (2002) brings together Bradford Law and information seeking behavior. . . . the key point is that the distribution tells us that information is neither randomly scattered, nor handily concentrated in a single location. Instead, information scatters in a characteristic pattern, a pattern that should have obvious implications for how that information can most successfully and efficiently be sought (Bates, 2002).

Bates applies conceptually different search techniques (directed searching, browsing and linking) to the Bradford zones. Bates postulates the Bradford nucleus for browsing, the second zone for directed searching with search terms and further zones for linking. We focus on an automatic change from directed searching (enhanced by treatment of semantic heterogeneity) into browsing. Starting with a subject specific descriptor search, we will connect the query with our heterogeneity service to transfer descriptor terms into a multi-database scenario. In the second step, the results from the different databases will be combined and sorted according to Bradford’s method (i.e. most productive journals for a topic first). The conclusion this step provides us with a Bradfordized list of journal articles. The next step is the extraction of a result set of all documents in the Bradford nucleus which can be delivered for browsing. This automatically generated browsing modus can be compared to Bates search technique ‘‘journal run’’. The focus on a re-ranking technique based on Bradfordizing is interesting owing to the universal properties of the law, allowing it to be applied in a one-database scenario (e.g. Mayr and Umsta¨tter, 2007) and a multi-database scenario like vascoda or sowiport. On a very abstract level, Bradford re-ranking can be used as a compensation method for enlarged search spaces; however, in our application model the information on the core journals is used for document ranking. Co-author networks It is generally acknowledged that standard search services do not meet the wealth of information material supplied by DLs. Traditional retrieval systems are strictly document oriented such that a user is unable to exploit the full complexity of the information stored therein. Bibliographic data, for instance, offers a rich information structure that is usually ‘‘hidden’’ in traditional information systems. A typical example of this issue are link structures among authors, given, for instance by coauthor relationships, and – more importantly – the strategic position of authors within a given collaboration structure. Moreover, relevant information is more and more distributed over several heterogeneous information sources and services (e.g. bibliographic reference services, citation indices, full-text services, etc.). DLs are therefore only meaningfully usable if they provide high-level search services that fully exhaust the information structures stored and, at the same time, reduce the complexity of information to highly relevant items. Due to the notion of a Semantic Web, particularly the Friend of a Friend approach[9], this strongly suggests the development of techniques that overcome the strict document orientation of standard indexing and retrieval methods by providing a deeper analysis of link structures and the centrality of entities in a given network structure. This approach focuses on network analysis concepts for extracting central actors in co-author networks and ranking documents by author centrality (Mutschke, 2003). The expressiveness of co-author networks has been demonstrated in a number of scientometric studies (see e.g. Beaver, 2004). The basic approach of our model is to reason about the network structure in order to evaluate relevant authors for a particular domain. This information on the centrality of authors within their scientific community is then used to rank documents. According to graph theory, a co-author network in our model is described as a graph G ¼ (V, E), where V is the set of vertices (authors), and E the set of edges (coauthorships). A co-author network is generated on the basis of all co-authorships that appear in a given document set (e.g. the result set of a query). On social networks a

Reducing semantic complexity 219

LR 57,3

220

number of calculations can be performed. An important structural attribute of the vertices is their centrality. Centrality measures the contribution of a network position to the vertices’ prominence, influence or importance within a social structure. In our model, we use the betweenness measure. Betweenness focuses on the ratio of shortest paths a vertex lies on. An author with a high betweenness is therefore a vertex that connects many authors in the network. Betweenness is therefore seen as a measure indicating an actor’s degree of control or influence of the interaction processes that construct a network structure. Accordingly, an index of centrality within a scientific collaboration and communication structure might indicate the degree of relevance of an author for the domain in question; such relevance would be attributable to his/her key position in the network. In our application model information on the centrality of authors is used for document ranking. This is done by weighting the documents retrieved by the centrality values of their authors such that the user is provided with documents of central authors. Figure 3 visualises the planned application of the value-added services (in the stages of the search process and the combination of the single components). See the STR in the beginning of a search and re-ranking of combined search result sets at the end of a search loop (see stages 2 and 7 in Figure 3). Integration Beyond an isolated use, a combination of the approaches is promising to yield much higher innovation potential. In our model, the following scenarios are supported (e.g. combining Bradfordizing with Author Centrality as in Figure 4). The user is provided with publications which are associated with both central authors as well as core journals. From a technical point of view, the following variants are suitable and may yield different results: .

Bradfordizing as a filter for the network analysis process: central authors are evaluated within the set of documents which are associated with core journals (i.e. the result set is reduced to the core journal set before author centrality analysis is performed).

.

Author centrality as a filter for Bradfordizing (the ‘‘inverse’’ version of the variant above): Bradfordizing is performed on the set of result set document which are assigned to central authors (i.e. the result set is reduced to ‘‘central’’ documents before core journals are evaluated).

.

The ‘‘intersection’’ variant: core journals and central authors are first evaluated independently from one another on the basis of the whole result set. Publications that satisfy both relevance criteria (they appear in a core journal and their authors are central) are determined in a second step (see Figure 4).

Those combination models could not only be applied to result set re-ranking but also to the search term recommendation process (i.e. the usage of Bradfordizing and author centrality analysis as a filter on the collection used for the STR analysis). Future research work should address the use of information pertaining to institutions, themes or citations as a means of providng further value-adding functions in re-ranking methods, rather than just using authors and journals. An important further research issue is to apply and evaluate the proposed ranking methods at the user search stage in order to improve the precision of the initial result set.

Reducing semantic complexity 221

Figure 3. Combination and embedding of the modules: STR and re-ranking

Outlook The central impact of the paper focuses on the integration of three structural valueadding methods which aim at reducing the semantic complexity represented in distributed DLs at several stages in the information retrieval process: query construction, search and ranking and re-ranking. The integration of the models will be done using Semantic Web technologies which should enhance further insights into the usage of these techniques. The intersection of the Semantic Web world with the DL world as mentioned in Krause (2008) (available in this issue of Library Review) will

LR 57,3

222

Figure 4. Intersection of core journal and central author documents

hopefully result in more sophisticated analytical tools and interfaces for the presentation of information adapted to users’ needs. Notes 1. Elsevier. ‘‘Scirus – for scientific information only’’. Retrieved October 2007, from www.scirus.com/. 2. Online Computer Library Center (OCLC). ‘‘WorldCat’’. Retrieved October 2007, from www.oclc.org/worldcat/. 3. Department of the Classics, Tufts University. ‘‘The Perseus Digital Library’’. Retrieved October 2007, from http://perseus.mpiwg-berlin.mpg.de/. 4. ‘‘vascoda - Entdecke Information’’. Retrieved October 2007, from www.vascoda.de/. 5. World Wide Web Consortium (W3C). (2001). ‘‘Semantic Web Activity’’. Retrieved October 2007, from www.w3.org/2001/sw/. 6. ‘‘Competence Center Modeling and Treatment of Semantic Heterogeneity’’. Retrieved October 2007, from www.gesis.org/en/research/information_technology/komohe.htm. 7. Results from a German terminology mapping effort: intra- and interdisciplinary crossconcordances between controlled vocabularies. Presented at the NKOS/ECDL Workshop in Budapest Hungary. Retrieved October 2007, from http:// dlist.sir.arizona.edu/2054/. 8. Sowiport. Retrieved October 2007, from www.sowiport.de. 9. Friend-of-a-Friend (FOAF). Retrieved October 2007, from http://xmlns.com/foaf/spec/. 10. For this variant a (configurable) threshold for centrality is needed.

References Bates, M.J. (2002), ‘‘Speculations on browsing, directed searching, and linking in relation to the bradford distribution’’, in Bruce, H., Fidel, R., Ingwersen, P. and Vakkari, P. (Eds.), Fourth International Conference on Conceptions of Library and Information Science (CoLIS 4), available at: www.gseis.ucla.edu/faculty/bates/articles/Searching_Bradford-m020430.html (accessed 20 December 2007). Beaver, D. (2004), ‘‘Does collaborative research have greater epistemic authority?’’, Scientometrics, Vol. 60 No. 3, pp. 309-408. Blair, D.C. (1990), Language and Representation in Information Retrieval, Elsevier Science Publishers Amsterdam, New York, NY, p. 335. Blair, D.C. (2002), The challenge of commercial document retrieval, part II: a strategy for document searching based on identifiable document partitions’’, Information Processing and Management, Vol. 38 No. 2, pp. 293-304. Blair, D.C. (2003), ‘‘Information retrieval and the philosophy of language’’, Annual Review of Information Science and Technology, Vol. 37, pp. 3-50. Bradford, S.C. (1948), Documentation, Lockwood, London, p. 156. Gey, F., Chen, H., Norgard, B., Buckland, M., Kim, Y., Chen, A., Lam, B., Purat, Y. and Larson, R. (1999), ‘‘Advanced search technology for unfamiliar metadata’’, Third IEEE Metadata Conference, Bethesda, MD. Hellweg, H., Krause, J., Mandl, T., Marx, J., Mu¨ller, M.N.O., Mutschke, P. and Stro¨tgen, R. (2001), ‘‘Treatment of semantic heterogeneity in information retrieval’’, IZ Working Paper; No 23, IZ Sozialwissenschaften, Bonn, p. 47, available at: www.gesis.org/Publikationen/Berichte/ IZ_Arbeitsberichte/pdf/ab_23.pdf (accessed 20 December 2007). Krause, J. (2006), ‘‘Shell model, semantic web and web information retrieval’’, in Harms, I., Luckhardt, H.-D. and Giessen, H.W. (Eds), Information und Sprache: Beitra¨ge zu Informationswissenschaft, Computerlinguistik, Bibliothekswesen und verwandten Fa¨chern, Festschrift fu¨r Harald H. Zimmermann,Saur, Mu¨nchen, pp. 95-106. Krause, J. (2007), ‘‘The concepts of semantic heterogeneity and ontology of the Semantic Web as a background of the German science Portals vascoda and sowiport’’, in Prasad, A.R.D. and Madalli, D.P. (Eds), International Conference on Semantic Web and Digital Libraries (ICSD 2007), Documentation Research and Training Centre, Indian Statistical Institute, Bangalore, pp. 13-24, available at: https://drtc.isibang.ac.in/bitstream/1849/307/1/ 002_p39_krause_germany_formatted.pdf (accessed 20 December 2007). Krause, J. (2008), ‘‘Semantic heterogeneity: comparing new Semantic Web approaches with those of digital libraries’’, Library Review, Vol. 57 No. 3. Larson, R.R. (1991) ‘‘Classification clustering, probabilistic information-retrieval, and the online catalog’’, Library Quarterly, Vol. 61 No. 2, pp. 133-73. Larson, R.R. (1992), ‘‘Experiments in automatic library-of-congress classification’’, Journal of the American Society for Information Science, Vol. 43 No. 2, pp. 130-48. Liang, A.C. and Sini, M. (2006), ‘‘Mapping AGROVOC and the Chinese Agricultural Thesaurus: definitions, tools, procedures’’, New Review in Hypermedia and Multimedia, Vol. 12 No. 1, pp. 51-62. Mayr, P. and Umsta¨tter, W. (2007), ‘‘Why is a new Journal of Informetrics needed?’’, Cybermetrics, Vol. 11 No. 1, available at: www.cindoc.csic.es/cybermetrics/articles/v11i1p1.html (accessed 20 December 2007). Mayr, P. and Walter, A.-K. (2007a), ‘‘Einsatzmo¨glichkeiten von Crosskonkordanzen’’, in Stempfhuber, M. (Ed.), Lokal - Global: Vernetzung wissenschaftlicher Infrastrukturen: 12, Kongress der IuK-Initiative der Wissenschaftlichen Fachgesellschaft in Deutschland, GESIS – IZ Sozialwissenschaften, Bonn, pp. 149-66, available at: www.gesis.org/Information/

Reducing semantic complexity 223

LR 57,3

224

Forschungsuebersichten/Tagungsberichte/Vernetzung/Mayr-Walter.pdf (accessed 20 December 2007). Mayr, P. and Walter, A.-K. (2007b), ‘‘Zum Stand der Heterogenita¨tsbehandlung in vascoda: Bestandsaufnahme und Ausblick’’, in BID (Ed.), Information und Ethik 3. Leipziger Kongress fu¨r Information und Bibliothek, Verlag Dinges and Frick, Leipzig, available at: www.opus-bayern.de/bib-info/volltexte/2007/290/ (accessed 20 December 2007). Mutschke, P. (2003), ‘‘Mining networks and central entities in digital libraries: a graph theoretic approach applied to co-author networks’’, IDA 2003 – The fifth International Symposium on Intelligent Data Analysis, Berlin, available at: http://fuzzy.cs.uni-magdeburg.de/confs/ ida2003/ (accessed 20 December 2007). Plaunt, C. and Norgard, B.A. (1998), ‘‘An association-based method for automatic indexing with a controlled vocabulary’’, Journal of the American Society for Information Science, Vol. 49 No. 10, pp. 888-902. Vizine-Goetz, D., Hickey, C., Houghton, A. and Thompsen, R. (2004), ‘‘Vocabulary mapping for terminology services’’, Journal of Digital Information, Vol. 4 No. 4, available at: http:// jodi.tamu.edu/Articles/v04/i04/Vizine-Goetz/ (accessed 20 December 2007). White, H.D. (1981), ‘‘‘‘Bradfordizing’’ search output: how it would help online users’’, Online Review, Vol. 5 No. 1, pp. 47-54. Zeng, M.L. and Chan, L.M. (2004), ‘‘Trends and issues in establishing interoperability among knowledge organization systems’’, Journal of the American Society for Information Science and Technology, Vol. 55 No. 3, pp. 377-95. Corresponding author Philipp Mayr can be contacted at: [email protected]

To purchase reprints of this article please e-mail: [email protected] Or visit our web site for further details: www.emeraldinsight.com/reprints

The current issue and full text archive of this journal is available at www.emeraldinsight.com/0024-2535.htm

Faceted infrastructure for semantic digital libraries

Semantic digital libraries

A.R.D. Prasad and Devika P. Madalli Documentation Research and Training Centre, Indian Statistical Institute, Bangalore, India Abstract

225 Received 22 October 2007

Purpose – The paper aims to argue that digital library retrieval should be based on semantic Reviewed 2 November 2007 Accepted 13 November 2007 representations and propose a semantic infrastructure for digital libraries. Design/methodology/approach – The approach taken is formal model based on subject representation for digital libraries. Findings – Search engines and search techniques have fallen short of user expectations as they do not give context based retrieval. Deploying semantic web technologies would lead to efficient and more precise representation of digital library content and hence better retrieval. Though digital libraries often have metadata of information resources which can be accessed through OAI-PMH, much remains to be accomplished in making digital libraries semantic web compliant. This paper presents a semantic infrastructure for digital libraries, that will go a long way in providing them and web based information services with products highly customised to users needs. Research limitations/implications – Here only a model for semantic infrastructure is proposed. This model is proposed after studying current user-centric, top-down models adopted in digital library service architectures. Originality/value – This paper gives a generic model for building semantic infrastructure for digital libraries. Faceted ontologies for digital libraries is just one approach. But the same may be adopted by groups working with different approaches in building ontologies to realise efficient retrieval in digital libraries. Keywords Digital libraries, Worldwide web, Information management Paper type Research paper

Introduction From a user perspective, any digital library is only as good as its retrieval efficiency. The usual resource discovery tools provided by digital libraries are the classic ‘‘browse’’ and ‘‘search’’ facilities. Digital libraries, in their early forms, carried forward or inherited the legacy of bibliographic databases and Internet search engines. Such systems extracted search terms, pulling them out of their context and hence seldom meeting with any user satisfaction (Levene, 2006). Digital libraries evolved to offer resources in domain specific, document type specific and user level/type specific environments. Some digital libraries are hybrids that contain heterogeneous contents and cater for diverse and distributed user communities. Digital libraries can thus be simple models having homogeneous on-disk resources where they just serve a set of users in one given environment. At the other end, digital libraries can be complex systems that are required to process hundreds of parallel requests in diverse scenarios. Metadata standards and resource discovery Metadata standards brought into focus the fact that there is a need to describe the content of Internet resources using some ‘‘meaningful descriptors’’ (Smith and Schirling, 2006). Metadata standards are implemented so as to provide much needed pointers to information on web by breaking up the content into bits and pieces. Metadata standards are particularly applicable to digital libraries as tools for finding information. The function of metadata directly corresponds to card catalogues

Library Review Vol. 57 No. 3, 2008 pp. 225-234 # Emerald Group Publishing Limited 0024-2535 DOI 10.1108/00242530810865493

LR 57,3

226

in traditional print libraries. The logical extension of such data description methods were carried forward by bibliographic databases. Data was structured into fields and subfields that followed standards in the elements set. The element sets were formalised into standards such as MARC21 (Library of Congress, 2006) and UNIMARC (IFLA, 2004). These standards were then widely adopted by the libraries into their OPACs. Also, there emerged several standards for the data formats themselves, including ISO date format, AACR2 rules for data extraction, and so forth. For the purpose of this discussion, we would like to broadly classify metadata into two kinds: bibliographic and non-bibliographic metadata. Bibliographic metadata is the kind that is based on traditional library catalogue data. Non-bibliographic metadata could be descriptions of any entity, such as a person, organisation, service, or market products, etc. Traditionally, machine readable catalogues based on standards like MARC21 include authority files for personal names, organisations and institutions, geographical places, etc. Mostly these authority files are used for vocabulary control or standardisation of terminology used within systems. In digital library environments though, such authority files can be transformed and enriched using metadata. In this paper we restrict ourselves to only bibliographic metadata. Metadata is much akin to bibliographic data in that it is ‘‘structured data’’ which describes the characteristics of a resource. Metadata is about knowledge, which is the ability to turn information and data into effective action (Haynes, 2004). This implies that through metadata, we can attribute more ‘‘contextual’’ information for the resources that would eventually help in their discovery by the end users on the web. The direct implication is that it empowers search engines or end users in searching by providing essential clues. For example, searching ‘‘Johnson’’ would produce the results where ‘‘Johnson’’ is a player, ‘‘Johnson’’ is a company (producing or marketing toothpaste to electric geysers!) or ‘‘Johnson’’ as a name of computer program. But if we were given a mechanism to state in the query, the following: .

look for ‘‘Johnson’’ where ‘‘Johnson’’ is the author, then we could only retrieve those resources where Johnson appears as author.

This was perfected in bibliographic databases on CDs and online information systems. In web parlance, metadata elements are used in place of bibliographic elements. But the above option would fail again if we are looking for, ‘‘Johnson’’ – an electrical geyser company. Obviously this requires a different kind of metadata other than just bibliographic. So each requirement is context specific. Dublin Core, the widely used standard for metadata, describes web pages structured into 22 elements that can be used optionally and each element can be used repetitively as well. Elements are qualified further by element refinements and data standards (Dublin Core Metadata Initiative, 2006). While the design and content of Dublin Core is very simple, the need was felt by different communities that it should be possible to extend or change it to suit the description of different datasets. However, Dublin core is used as minimum common denominator among and between digital libraries. A plethora of metadata standards have followed to suit the needs of different types of content, domains, and services. A few are mentioned below: (1) Particular domains: AGRIS AP (agricultural metadata standard). (2) Document type collection: ETD-MS (electronic theses and dissertations). (3) Sector specific: GILS (Government information). (4) Services specific: LOM (eLearning).

Whatever the type of metadata or whatever the purpose to which they are applied, DLs use the given descriptive metadata to get clues about the data within each element. The emergence of the web meant that metadata standards had to be carried in a language processable by Internet tools such as browsers, search engines or other webbased information applications, and this led to eXtensible Markup Language (XML) applications for metadata. XML and domain specific markup The advent of XML ushered in the possibility of extending element sets that were in HTML, limited mostly to rendering and formatting commands. XML enabled the web resource providers to describe their content in a set of elements according to their own needs (Castle, 2006; Bergholz, 2000). This implied that the extended elements sets could now include elements to indicate data contained within them. One of the first applications of XML was hence in the representation of library data, though the role of the library community in evolving metadata standards was only acknowledged much later. Bibliographic data and digital library records could easily be represented in XML (see, for example MODS, 2007; MARCXML, 2007). Many communities depending upon their subject domains or activities designed specific element sets to describe the content of their resources resulting in many domain specific and community specific XML applications. The presence of many such sets of XML-based metadata only posed more problems for retrieval. The existence of varied elements (even in a given domain) with different elements sets, as well as synonymous and homonymous XML names. The initial excitement associated with the delivery of diverse XML applications soon brought to the fore the usual problems in search and retrieval. If the aim is to achieve efficient and precise retrieval, it is insufficient to merely have descriptive metadata. It is necessary to encode the semantics of resource, along with the context of the concepts in the content. Moreover, the semantics should be not only for human understanding, but also for machine understanding (i.e. processable by computers or intelligent applications). This brings into sharp focus the approaches to resource discovery deployed by libraries to library collections, such as subject based indexing techniques using controlled vocabularies. Much the same is being tried on the web by resource description framework (RDF) and web ontology language (OWL) representations that attempt to encode concepts as well as conceptual relations. Semantic Web technologies There is enormous amounts information on web, much of which is probably very useful to many of its users. The only problem is that it seldom reaches the right users at the right time and in the right form. This problem is because search engines (and other tools used to access information on web) have little idea of the meaning of terms in the context they are sought. The information on web is now for human-to-human communication, not for human-machine-human communication. If the content on web was represented such that machines could manipulate it meaningfully, then the retrieval would be more meaningful or ‘‘semantic’’. This is the underlying principle of Semantic Web (Berners-Lee et al., 2001). The main goal of Semantic Web is to develop languages for expressing information in a machine processable way (Dumbill, 2001). The Semantic Web has presented many tools and techniques, such as the RDF[1] and OWL[2]. These tools can be enriched with the variable data and content which

Semantic digital libraries

227

LR 57,3

228

Figure 1. Evolution towards semantic representation

library professionals have developed in the form of cataloguing schemes like AACR2, MARCs, authority files (personal, institutional, organisational, etc.) in the case of metadata; and classification schedules and thesauri in the case of ontology. However, it is not the simple transformation of librarians’ tools to the web technology alone that enables the provision of web based information services. The W3C has ratified a set of web languages to express the meaning of the information sources on the web. These set of languages are meant to represent the meaning of information in structured way. These languages deal with different layers of expression when dealing with semantics. This set of languages and technologies comprise of XML, XML Schema, RDF, RDF Schema and OWL, and Simple Knowledge Organisation Systems (SKOS)[3]. SKOS has been developed as a RDF-based language to represent various kinds of knowledge organisation systems such as taxonomies, classification schemes, thesauri and other controlled vocabulary schemes. It is a framework for expressing knowledge organisation systems in a machine-understandable way. SKOS Core provides a model for expressing the basic structure and content of concept schemes (Miles and Brickley, 2005). As SKOS is based on the RDF model, it assigns URIs to all concept-terms of the concept schemes. Though there are peculiarities between them (e.g. RDF is not based on XML), each of these languages demonstrates a degree of evolution in our ability to represent semantic data on the web (Figure 1). Based on such explicit representation, inferencing has to be made to produce information services and products as required by end-users. Information represented in these languages is further processed by various inference engines. The precision of inferencing always depends upon the extent to which the information is represented, encoding both the meaning of concepts and their relation with other concepts. Ontology is the preferred knowledge representation tool in the Semantic Web, through domain knowledge can be expressed in machine processable way. In the Semantic Web community the most widely quoted definition of ontology is that given by Gruber (1993): ‘‘an explicit specification of a conceptualisation’’. Noy and McGuinness (2001) have defined ontology as a ‘‘common vocabulary through which the information of a domain can be shared’’. This includes ‘‘machine interpretable definitions of basic concepts in the domain and relations among them’’. In both the above definitions of ontology it is clear that by ontology we express the basic concepts of a domain, and make the relationships between these concepts explicit. Thus, ontologies are essentially formal conceptualisations of domains. By

making the concepts and relationships explicit, we can make the information on the web conducive to machine processing. Semantic digital libraries It is not just the words/terms that appear in a work that represent the document, it is the semantics of the document that is behind the words that represents the document. It may be the case sometimes that a word that did not occur in a document may be one of the potential representative or surrogates of a document. Conversely, many words that appear in a document may not be candidates for representing the semantics of the document. The occurrence of some keywords, or the sum of some keywords, may not represent the meaning. The LIS field has a long history of handling voluminous information and the basic tools of information retrieval have been cataloguing and classification. Cataloguing (metadata) describes information embedded in a document whether printed or digital, and classification (ontology) establishes the relations between documents. If we wish to organise information on the web for providing various web-based information services, it goes without saying that we require both metadata and ontology. One problem with present search engines is that they are uniterm based (i.e. in library science parlance, based on what is known as post-coordinate indexing). The end user may use some Boolean operators, combining the words/phrases that occur in web pages. The problem with uniterm indexing is that the words are forcefully taken out of their context and placed in the index without context. LIS researchers have proposed context-sensitive indexing methods like PRECIS (Austin, 1974) and POPSI, (Bhattacharyya, 1981) which basically do pre-coordinate indexing. That is, words along with their semantic context are represented in the index. These pre-coordinate indexing techniques appeared to be emulated in a machine environment. Now, with the latest developments in AI, web ontologies and the Semantic Web, it is time to reexamine these pre-coordinate or context-sensitive indexing systems in order to achieve greater precision in web and digital library retrieval, however costly they appear to be when compared to the coarse statistical techniques of relevance ranking. Semantic infrastructure for digital libraries Digital library system architectures include either a top down or a bottom up approach. In the first instance the systems are built according to planned and anticipated services and in the second as a store for assets to be manipulated and processed in order to serve some form of output as desired. These approaches correspond to the ‘‘user centric’’ and ‘‘machine/item centric’’ approaches (Maly et al., 1999; French and Viles, 1999; Jane`e and Frew, 2003; Li and He, 2007). The problem with user centric models is that because they begin by providing solutions to user requirements, they cannot generically give solutions to all user information seeking approaches. Even the so-called faceted approach, as followed by Flamenco (Flamenco project, 2004), only allows users to ‘‘drills down’’ in a single concept path; whereas the user may want not only the concept but the relationships encoded with respect to its relative distances from another given concept. Centre-out approach It is often the purpose that guides system design; yet, specificity can be lost in such generalised approaches. This is the classic trade-off in the design of knowledge discovery systems. Ideally, we should have generalised systems that directs the user to

Semantic digital libraries

229

LR 57,3

230

more and more specific systems; perhaps a single system may not provide both broad and specific information requirements (e.g. consider the differing information requirements of a general doctor and a specialist surgeon). To this end, we propose a centre-out approach to achieving semantic digital libraries; the centre being a semantic representation based on faceted ontologies of information in a generic form. The knowledge structure corresponds to a semantic framework that is really a generalisation or abstraction of a formal representation of subject domains. Each domain is envisaged as consisting of divisions or pieces of knowledge in facets where a facet is a distinctive division of the domain or a subject that is conceptualised. Each facet itself contains a set of concepts of the domain in a hierarchy, and many such facets together comprise a subject domain. This model is based on Ranganathan’s theory of facetisation (Ranganathan, 1967). The generic manifestations of such subject representations lead to faceted ontologies. A detailed account of facetisation is outside the scope of this paper. Needless to say, such facetisation lays out a subject according to different perspectives so that users can pick and choose the components as required. Faceted ontologies form the central conceptualisations of domains contained in digital libraries, providing semantic services through a layer of inferencing. Hence we propose a facet representation system as illustrated by Figure 2. Semantic store The semantic store corresponds to a knowledge database of semantically segregated facets of domains relating to the digital library content. The store involves encoding term meanings, synonyms, variations, linguistic clues, and so forth (Figure 3). Knowledge structure The knowledge structure is a formalised hierarchical structure of concepts for each facet. All such facets are assembled together as required by each scenario. The facets are recognised as entities, action and properties of the entities identified (Bhattacharyya, 1981) (Figure 4).

Figure 2. Workflow of centre-out approach of digital library model

Semantic digital libraries

231

Figure 3. Semantic store

Subject representations/surface structures While the facets themselves are distinct divisions of the domain and contain the concepts belonging to the facet within them, there are rules to generate surface strings for any given term that brings along with it the context. These are formalised strings that represent concepts in their entire context, generated for tracing a term or its variant form from a specific domain(s). They are formalised by using certain rules in the system about how each term is related to the other. For example (Figure 5), we project harvesting of sugarcane in India in the following strings based on the faceted ontologies that underlies the digital library.

Figure 4. Knowledge structure

Figure 5. Subject representations

LR 57,3

232

Inferencing The inferencing mechanism attempts to generate a list of required services based on the information requirements stated by users. The system receives the user query, and analyses and normalises the linguistic part by recognising synonyms, word variants and homonyms. Ideally, it presents to the user all the possible paths to a given query. The user would be free to modify and re-formulate the query based on the suggestions provided by the system. (User query reformulation is not the focus of this paper and will be discussed as part of the findings of user interaction experiments, to be published at a later date). Further, the inferencing mechanism matches the refined user query (represented by terms) to the semantic faceted representation of the relevant domain in the semantic store. Services/products Services and products are generated corresponding to distinct scenarios and user requirements. This stage could be interactive essentially helping in decision making and paths to be taken by the inferencing mechanism. A typical example might relate to search result display. Conclusion The natural course of digital libraries is towards the Semantic Web. Digital libraries are the best candidate systems for experiments deploying Semantic Web technology. In the first instance, digital libraries have the structured data necessary to realise the Semantic Web vision; digital libraries are empowered with metadata, while metadata remains a rarity with respect to general Web resources. Moreover, digital library content is scope specific, and it is easier to build representation systems such as faceted ontologies for the content therein. The content of digital repositories, whether it is books or articles or even e-learning modules, have to be related and presented to the machine in order to be processed and interpreted. Classically, libraries had systems that processed subjects or domains and built representations, such as subject indexed strings. These strings had enough contextual information (using the method of facetisation and synthesis) that they formed a semantic formalisation of the domain scope of library collections. The semantic model proposed here is an emulation of the same function of libraries. The dimensions of course have changed to the digital online world of digital libraries. Metadata alone will not suffice to achieve precision in retrieval. It is semantic systems – with faceted knowledge representation at its core – that will project the content of the digital libraries in their entirety so that users can enter the system at an appropriate point to match their information need. Implementing and evaluating our proposed faceted infrastructure model is the focus for future research work, the outcome of which will be documented in the literature in due course. Notes 1. See www.w3.org/RDF (accessed 20 December 2007). 2. See www.w3.org/2007/OWL/wiki/OWL_Working_Group (accessed 20 December 2007). 3. See www.w3.org/2004/02/skos/ (accessed 20 December 2007).

References Austin, D. (1974), Precis: a Manual of Concept Analysis and Subject Indexing, Council of the BNB, London. Bergholz, A. (2000), ‘‘Extending your markup: an XML tutorial’’, IEEE Internet Computing, Vol. 4 No. 4, pp.74-9. Berners-Lee, T., Hendler, J. and Lassila, O. (2001), ‘‘The semantic web’’, Scientific American, Vol. 284 No. 5, available at: www.sciam.com/article.cfm?articleID=00048144-10D2-1C7084A9809EC588EF21 (accessed 20 December 2007). Bhattacharyya, G. (1981), ‘‘Subject indexing language: its theory and practice’’, in Proceedings of the DRTC Refresher Seminar – 13, New Developments in LIS in India, DRTC, Bangalore. Castle, D. (2006), ‘‘Comparative evaluation of XML information retrieval systems’’, Proceedings of the 5th International Workshop of the Initiative for the Evaluation of XML Retrieval, INEX 2006, Springer Verlag, Berlin. Dublin Core Metadata Initiative (2006), Dublin Core Metadata Element Set, Version 1.1, available at: http://dublincore.org/documents/dces/ (accessed 20 December 2007). Dumbill, E. (2001), The Semantic Web: A Primer, XML.com, 01 November, available at: www.xml.com/pub/a/2000/11/01/semanticWeb/index.html (access 20 December 2007). Flamenco Project (2004), ‘‘The Flamenco search interface project’’, available at: http:// flamenco.berkeley.edu/index.html (accessed 20 December 2007). French, J.C. and Viles, C.L. (1999), ‘‘Personalized information environments: an architecture for customizable access to distributed digital libraries’’, D-Lib Magazine, Vol. 5 No. 6, available at: www.dlib.org/dlib/june99/french/06french.html (accessed 20 December 2007). Gruber, T.R. (1993), ‘‘A translation approach to portable ontology specification’’, Knowledge Acquisition, Vol. 5 No. 2, pp. 199-220. Haynes, D. (2004), Metadata for Information Management and Retrieval, Facet Publishing, London. IFLA (2004), ‘‘UNIMARC – a brief overview’’, available at: www.unimarc.net/brief-overview.html (accessed 20 December 2007). Jane´e, G. and Frew, J. (2003), ‘‘The ADEPT digital library architecture’’, in Proceedings of the Joint Conference on Digital Libraries (JCDL 2002), 13-17 July, Portland, Oregon, available at: www.dpi.inpe.br/cursos/ser303/jcdl-adept.pdf (accessed 20 December 2007). Levene, M. (2006), An Introduction to Search Engines and Web Navigation, Addison Wesley, NY. Li, J. and He, J. (2007), ‘‘User-centric model for supporting web services’’, in Proceedings of the International Multi-Conference on Computing in the Global Information Technology, IEEE Computer Society, Washington DC. Library of Congress (2006), ‘‘MARC 21 concise format for bibliographic data’’, available at: www.loc.gov/marc/bibliographic/ (accessed 20 December 2007). Maly, K., Nelson, M.L. and Zubair, M. (1999), ‘‘Smart objects, dumb archives: a user-centric, layered digital library framework’’, D-Lib Magazine, Vol. 5 No. 3, available at: www.dlib.org/dlib/march99/maly/03maly.html (accessed 20 December 2007). MARCXML (2007), ‘‘MARC21 XML schema’’, available at: www.loc.gov/standards/marcxml/ (accessed 20 December 2007). Miles, A. and Brickley, D. (2005), ‘‘SKOS core guide: W3C working draft 2 November 2005’’, available at: www.w3.org/TR/2005/WD-swbp-skos-core-guide-20051102/ (accessed 20 December 2007). MODS (2007), ‘‘Metadata object description schema’’, available at: www.loc.gov/standards/mods/ (accessed 20 December 2007).

Semantic digital libraries

233

LR 57,3

234

Noy, N.F. and McGuinness, D.L. (2001), ‘‘Ontology development 101: a guide for creating your first ontology’’, available at: http://protege.stanford.edu/publications/ontology_development/ ontology101.html (accessed 20 December 2007). Ranganathan, S.R. (1967), Prolegomena to Library Classification (Ranganathan Series in Library Science, 20), Asia Publishing House, London. Smith, J.R. and Schirling, P. (2006), ‘‘Metadata standards roundup’’, IEEE Multimedia, Vol. 13 No. 2, pp. 84-8. Further reading Gonc¸alves, M.A., Moreira, B.L., Fox, E.A. and Watson, L.T. (2007), ‘‘What is a good digital library: a quality model for digital libraries’’, Information Processing & Management, Vol. 43 No. 5, pp. 1416-37. Corresponding author A.R.D. Prasad can be contacted at: [email protected]

To purchase reprints of this article please e-mail: [email protected] Or visit our web site for further details: www.emeraldinsight.com/reprints

The current issue and full text archive of this journal is available at www.emeraldinsight.com/0024-2535.htm

Semantic heterogeneity: comparing new semantic web approaches with those of digital libraries Ju¨rgen Krause GESIS-IZ, Bonn and Computer Science Department, University of Koblenz-Landau, North Rhine-Westphalia, Germany

Semantic heterogeneity

235 Received 19 October 2007 Reviewed 19 October 2007 Accepted 20 November 2007

Abstract Purpose – To demonstrate that newer developments in the semantic web community, particularly those based on ontologies (simple knowledge organization system and others) mitigate common arguments from the digital library (DL) community against participation in the Semantic web. Design/methodology/approach – The approach is a semantic web discussion focusing on the weak structure of the Web and the lack of consideration given to the semantic content during indexing. Findings – The points criticised by the semantic web and ontology approaches are the same as those of the DL ‘‘Shell model approach’’ from the mid-1990s, with emphasis on the centrality of its heterogeneity components (used, for example, in vascoda). The Shell model argument began with the ‘‘invisible web’’, necessitating the restructuring of DL approaches. The conclusion is that both approaches fit well together and that the Shell model, with its semantic heterogeneity components, can be reformulated on the semantic web basis. Practical implications – A reinterpretation of the DL approaches of semantic heterogeneity and adapting to standards and tools supported by the W3C should be the best solution. It is therefore recommended that – although most of the semantic web standards are not technologically refined for commercial applications at present – all individual DL developments should be checked for their adaptability to the W3C standards of the semantic web. Originality/value – A unique conceptual analysis of the parallel developments emanating from the digital library and semantic web communities. Keywords Worldwide web, Digital libraries, Modelling, Information management Paper type Conceptual paper

Alternative concepts to web search engines like Google: Semantic foundations and heterogeneity In order to improve content indexing beyond that of Web search engines such as Google, one scientific approach is increasingly being discussed: the Semantic Web based on ontologies. The Semantic Web discussion (see for an overview, Fensel, 2001; Stuckenschmidt and Harmelen, 2005; Staab and Studer, 2004) began by focusing on the weak structure of the web and the lack of consideration for semantic information in content indexing. The research groups working on the Semantic Web maintained from the very beginning that the belief information retrieval models used by commercial search engines would lead to acceptable results – using automatic indexing of the websites in conjunction with an intelligent usage of hyperlinking structures – was not justified. Both groups began primarily from the ‘‘visible web’’. This was then further expanded with other sources of the ‘‘invisible web’’ (e.g. subject-specific databases, repositories, knowledge bases for companies, etc.). The points criticised by the Semantic Web and ontology approaches match those of the ‘‘Shell model approach’’ (Schalenmodell, Krause, 2006; 2007). The Shell model, which appeared in the mid-1990s, had an emphasis on the centrality of its heterogeneous

Library Review Vol. 57 No. 3, 2008 pp. 235-248 # Emerald Group Publishing Limited 0024-2535 DOI 10.1108/00242530810865501

LR 57,3

236

components. The argument of the Shell model did not begin with the web, but demanded restructuring and new research approaches for digital libraries and specialised information providers. In fact, the argument for the Shell model began with the ‘‘invisible web’’, which would then open to the visible web. In this case the actual challenge was that the paradigm of homogenisation through standardisation was partially sacrificed or was to be at least supplemented by intelligent processes for handling heterogeneity[1]. Here, just like with the Semantic Web approach, heterogeneity of semantic content analysis and indexing by various groups of providers was a central topic. The Shell model and its ‘‘heterogeneity components’’ represents a general framework in which specific types of documents with differing content indexing can be analysed and algorithmically related (Krause, 2006). Key to this are intelligent transfer components between the different types of content indexing (a special form of cross walk or mapping) that can accommodate the semantic differences. They conceptually interpret the technical integration between individual databases with differing content indexing systems by relating the terminologies of the domain-specific and general thesauri, classifications, etc. to each other. Semantic Web and ontologies According to Fensel (2001, p. VI), an ontology is ‘‘a community mediated and accepted description of the kinds of entities that are in a domain of discourse and how they are related’’. Stuckenschmidt and Harmelen (2005, p. IX) define the problem to be solved by ontologies as ‘‘information sharing’’. This encompasses the integration of heterogeneous information sources and results. [T]he problem of information sharing . . . is only solvable by giving the computer better access to the semantics of the information . . . we not only need to store such obvious metadata as its authors, title, creation date, etc., but we must also make available in a machine accessible way the important concepts that are discussed in the documents, the relation of these concepts with those in other documents, relating these concepts to general background knowledge.

Clearly information and documentation centres and libraries with their classifications, standardised authority files and thesauri, have a long tradition of characterising the intellectual content of documents. Consequently Umsta¨tter and Wagner-Do¨bler (2005, p. 54) argue: Ontologies in libraries, information science, and computer science are thesauri in which the basic meanings of word fields and their relations to one another are depicted in computers.

Hahn and Schulz (2004, p. 134), who take medical thesauri as a starting point for constructing ‘‘true’’ ontologies, clearly state that advocates of ontologies generally view these relations as insufficient: [UMLS Metathesaurus] [Their] semantics are shallow and entirely intuitive, which is due to the fact that their usage was primarily intended for humans . . . there is no surprise that the lack of a formal semantic foundation leads to inconsistencies, circular definitions, etc. . . . This may not cause utterly severe problems when humans are in the loop [as] its use is limited . . . [to] document retrieval tasks. Anticipating its use for more knowledge-intensive applications such as medical decision making . . . those shortcomings might lead to an impasse.

Ontologies should therefore enable more than just the homogenisation of search terms during database searching, which is why they require the possibility of automatic deductive processes.

Ontologies are formal structures supporting knowledge sharing and reuse (Fensel, 2001, p. 1). An ontology provides a domain theory and not the structure of a data container. In a nutshell ontology research is database research for the twenty-first century, where data needs to be shared and does not always fit into a simple table (Fensel, 2001, p. 10).

Heterogeneity among multiple ontologies Given the aforementioned ontological comments and definitions, it is clear that it is the formal deductive processes that solve the problem of heterogeneity between various different information sources. This was a central component of the ontology approach for the Semantic Web from the very beginning. Local models must be interwoven with other models, such as the social practice of the agents that use ontologies to facilitate their communication needs . . . We no longer talk about a single ontology, but rather about a network of ontologies. Links must be defined between these ontologies and this network must allow overlapping ontologies with conflicting – and even contradictory – conceptualizations (Fensel, 2001, p. 4).

Stuckenschmidt and Harmelen (2005) discuss two approaches for carrying out information sharing among multiple ontologies in a network: .

Each information domain utilises its own vocabulary and develops an ontology (the multiple ontology approach). There is no arrangement regarding a common vocabulary or a minimal global ontology. Stuckenschmidt and Harmelen (2005, p. 33) clearly reject this approach:

In reality the lack of a common vocabulary makes it extremely difficult to compare different source ontologies . . . the mapping has to consider different views of a domain . . . the mapping is very difficult to define, because of the many semantic heterogeneity problems which may occur .

Hybrid approaches allow different ontologies for different information domains, but are based on a jointly utilised (and agreed upon) vocabulary, which facilitates deductive mapping. The drawbacks are equally obvious:

The drawback of hybrid approaches, however, is that existing ontologies cannot be re-used easily (Stuckenschmidt and Harmelen, 2005, p. 34).

Problem areas from the perspective of information and documentation Just like the Shell model, ontologies in the framework of the Semantic Web focus on the inevitable heterogeneity that arises when attempting to exploit different information sources, a difficulty that needs to be overcome. They also emphasise the value of an indepth semantic indexing in comparison to approaches used by general web search engines. Two observations are important when making comparisons to the Shell model and its heterogeneity components: .

First, the difference between ontologies and thesauri is not so much about the concept, but about indexing depth.

.

And second, the semantic foundation of content analysis and indexing improves and brings with it the hope for better search results.

Knorz and Rein (2005, p. 1) illustrate the dilemma of this approach: This hope can be justified but not very well verified. Usages of ontologies can be extensively ordered along two ends of a spectrum . . . these usages of ontologies cover either a broad area

Semantic heterogeneity

237

LR 57,3

238

and work with semantic, barely differentiated relations, or they work with respect to attributes and relations in a very differentiated way with simultaneous limitations in a miniworld. In the first case the result is barely better than what a conventional thesaurus delivers, in the second case . . . the essential result can be obtained via conventional databases.

Multiple ontologies are specifically allowed in the Semantic Web. This means that the homogeneity paradigm of the LIS community to enforce standardisation through homogeneity appears to have been overcome. However, preferring the hybrid approach indicates that, instead, ontology modelling relies ultimately on a common vocabulary and defined relations between both terminological systems. In turn, this means that the direct utilisation of the invisible web – used as a basis up to now – is problematic, at least in the specialised science context. The Shell model and bilateral transfer modules Essentially then, ontologies are also focused on standardisation. This standardisation assumes a new form by considering various perspectives to be integrated. Principally, they try to do the same as the centralist approaches of the seventies in the information and documentation domain, but on a different level. The models of both groups coexist and agree on cooperation, and do so without means for hierarchical implementation. The classic demand for comprehensive standardisation makes sense and is, per se, not wrong: No heterogeneous components are needed if everyone uses the same thesaurus or the same classification. As long as everyone knows that efforts at standardisation can be only partially successful, then everything speaks in favour of initiatives such as these. But, no matter how successful they are in one subject area, the remaining heterogeneity, for example, with respect to various types of content analysis and indexing (e.g. automatic, different thesauri, various classifications, differences in categories included.) would be too great to ignore. The broad access to the web speaks against a centralist doctrine of content indexing per se. The question then remains for both approaches – for the Shell model and for ontologies. In contrast to ontology research, the Shell model emphasised the usage of existing semantic knowledge from the very beginning. Thesauri and classifications have been further refined over decades and connected through the process of intellectual indexing, specifically with the high quality information sources of the ‘‘invisible web’’. Their intelligent usage promises, in the mid-term, the greatest progress vis-a`-vis those of Google, or even traditional specialised databases and library catalogues. As mentioned in the introduction, the Shell model and its heterogeneity components represents a general framework in which specific types of documents with differing content indexing can be analysed and algorithmically related. Intelligent transfer components between the different types of content indexing (a special form of cross walk or mapping) that can accommodate the semantic differences are central to its operation. They conceptually interpret the technical integration between individual databases with differing content indexing systems by relating the terminologies of the domain-specific and general thesauri, classifications, etc. to each other. So far, this approach of handling semantic heterogeneity has been mainly implemented with the German science portal vascoda (www.vascoda.de) and the social science portal sowiport (www.sowiport.de), developed at the GESIS information centre in Bonn, Germany. vascoda is currently the most important project for achieving a new and innovative infrastructure in the field of scientific information in Germany. This new science portal merges controlled and qualitative collections from more than 40 providers. Its aim is to

integrate high-quality information from both the deep and the visible web by using search engine technology (FAST) and new concepts to integrate the data, not just technically, but to also solve the problem of semantic heterogeneity for achieving a high level of quality. Conceptually, the science portal vascoda is constructed upon two building blocks. These include a governing science portal, and the relatively independent-acting specialist portals of the individual academic subjects. Their construction also entails an additional problem area connected to that of semantic heterogeneity. Building up specialist portals like sowiport (for the social sciences) can be viewed as a multi-level process. It integrates national and international information of different types (metadata and full-text) and makes it available prepared for retrieval. This system offers the possibility for electronic publishing and discourse activities (i.e. discussion components), which is expanded to communication platforms via the search portals. In the long-term, this should lead to new forms of and higher quality of scientific working (Krause, 2007; Depping, 2007; Stempfhuber, 2006). Two different types of transfer modules have been implemented in sowiport and vascoda: .

Cross-concordances: the different terminology systems of classifications and thesauri are analysed in their context of usage and the terminologies are intellectually mapped to one another. This concept is not to be confused with metathesauri. Cross-concordances of the Shell model only contains those parts of the vocabulary where there are general semantic connections. A lot of terms can therefore remain unrelated. Crossconcordances only cover the static part of the transfer problem. This also differentiates them from ontological approaches.

.

Quantitative-statistic approaches: the transfer problem can generally be modelled as a vagueness problem between two content description languages. Various methods (e.g. probability methods, fuzzy approaches, rough set theory and neural networks) were suggested for the vagueness in information retrieval between user terminology and the database content that were also used for the transfer problem.

What is essential is that the transfer modules bilaterally operate on the database level (see Krause, 2004 for more details), connecting the different content description terms. None of the approaches carries the burden of transfer alone. They are entwined with one another and act in unison. This is both conceptually, as well as in practice, somewhat different from the traditional handling of vagueness between the user query on the one side and the document content of all databases on the other. A transfer module (e.g. V2) can be applied bilaterally between, for example, the content of a document that was indexed with a general key word list – as with SWD[2] – and a second document whose indexing is based upon a domain-specific thesaurus (like the social science thesaurus produced by the GESIS Social Science Information Centre (GESIS-IZ)) through qualitative processes such as cross-concordances and/or by another type of transfer module (Figure 1). The search algorithm can then establish a connection to the user terminology (V1) with a probabilistic method (Figure 2). In comparison to current information retrieval solutions, the important distinction is the possibility of using a specific type of transfer according to circumstances, and not just dealing with the problem of different terminology systems in an undifferentiated fashion.

Semantic heterogeneity

239

LR 57,3

240

Figure 1. Bilateral transformation

Bilateral transfer modules are – from a model building perspective – a very simple and basic building block that can quickly become very complex. The interplay of the transfer modules can be easily analysed and managed in the context of previous applications as is currently the case with sowiport or vascoda. Things can be expected to change quickly if the huge number of variations found on the visible and invisible web are taken into account. For this reason, the suggested model also needs more abstract modules that operate on a higher level of integration. The Shell model should accomplish this by complementing the bilateral transfer modules with a number of additional assumptions. For example, different levels of content indexing and document relevance are integrated into shells that are interconnected with each other through higher level transfer modules. The concept of bilateral transfer modules is now far enough advanced that it can be applied practically and promising initial empirical results are available (Krause, 2006).

Figure 2. Subjects and controlled vocabularies connected (by Philipp Mayr, GESISIZ, Bonn)

Semantic heterogeneity

241

Figure 3. Heterogeneity Service (HTS, by Philipp Mayr, GESIS-IZ, Bonn)

In the context of the GESIS–IZ project ‘‘competence centre modelling and treatment of semantic heterogeneity’’ (financed by the Federal Ministry of Education and Research, BMBF), 64 cross-concordances (including 513,000 term relations, based upon 25 controlled vocabularies) were developed (for more details see www.gesis.org/en/ research/information_technology/komohe.htm, Mayr and Walter, 2007a, b) between 2004 and 2007. The project is the largest terminology mapping effort in Germany. A heterogeneity service (HTS) was built up and implemented in sowiport using the social science cross-concordances. The HTS will be integrated into vascoda over the next six months. The service supports two scenarios: direct transformation of submitted user terms into equivalence relations and the presentation of the additional relations (of the cross-concordances) for users (Figure 3). The HTS translates between the various content indexing vocabularies. This system currently functions only if the vocabularies are interconnected with one another via cross-concordances. Users oftentimes choose their terms freely, meaning that these are either not, or only accidentally, part of controlled vocabularies. For this reason the improvement of an added service has been developed that reformulates a users’ freely chosen terms into a term(s) from a suggested controlled vocabulary. This so-called search term recommender increases hit accuracy (Mayr et al., 2008). Another addition is a complementary service, simultaneously employing two procedures borrowed from scientometry and network analysis. This service allows the re-ranking of the results list according to structural criteria. This entailes ranking the hit list according to core periodicals (so-called Bradfordising), and ranking according to the author’s significance within author networks (Mayr et al., 2008). The exceptional quality of these complementary services is gained through their connection to the other components. They specifically focus both on the search and the results (re-ranking), positively influencing one another[3]. The following section demonstrates that both approaches – the ontologies of the Semantic Web and the bilateral transfer module of the Shell model – supplement rather than exclude each another.

LR 57,3

242

The Shell model and its heterogeneity transfer modules vs ontologies of the Semantic Web Both the Shell model and the Semantic Web approaches acknowledge the weaknesses of the content analysis and indexing procedures employed by general web search engines. Both note the semantic foundation of information and its heterogeneity, and accept that heterogeneity is an acknowledged component of information sharing. In principle there is no conceptual discrepancy regarding the deeper semantic indexing which the ontologies aspire to. The German language example of searching for ‘‘Befinden von Kindern berufsta¨tiger Mu¨tter’’ (well-being of children of working mothers) in infoconnex (www.infoconnex.de) already poses a classic difficulty in the literature search. It reflects the criticism that thesauri only yield very limited relations, predominantly broader terms, narrower terms, similar terms and synonyms. Most references found with the above mentioned search terms yield documents dealing with the well-being of working mothers, not on the well-being of the child. The irrelevant hits can only be avoided through a more comprehensive explanation of the complex relations between the terms used in the thesauri. Since this would lead to a dramatic increase in the indexing work, thesauri usually do not contain these more complex relations. The underlying thesis is that the precision gained in some cases does not justify the increased effort. The conceptual background to information retrieval in this context is that the semantics will always remain unanalysed with regard to users’ search terms and descriptors for indexing. In contrast to that, ontologists require a formal semantic foundation with the possibility of formal deductive processes. Advocates of currently used thesauri argue that human intelligence supplements these non-analysed parts in man-machine interactions (Kuhlen, 2004)[4]. The question in traditional information retrieval and with regard to the Shell model is not whether parts of the semantics remain unanalysed, but rather whether the (intelligent) user is able to supplement these parts with acceptable (minor) effort when searching. In Krause and Stempfhuber (2005), one finds an indication for an increased semantic foundation with respect to the development of a social science portal, specifically relating to the ‘‘text-fact’’ integration of social science survey data and literature documents. In other words, there was a strong indication that users are typically unable to compensate for the missing semantic analysis when dealing with survey data rather than text. The fact is that the necessary compensation by users’ intelligence and research context fails, and not just in individual cases: It is the rule, rather than the exception. This example demonstrates that, in principle, no conceptual discrepancy exists between the bilateral transfer components of the Shell model and the ontology approach. The first may be interpreted as the lower level of an ontology, with reduced requirements regarding the depth of semantic indexing and limited deductive capabilities. The theoretical basis for such an approach is the aforementioned thesis; that within information retrieval contexts semantics may remain unanalysed since it is assumed that they can be compensated by user (human) intelligence. The natural language – partially not understood by the machine – serves, in this case, as a transport medium. Based on this, the semantic knowledge of thesauri and classifications may be used with relatively simple procedures and with the help of bilateral heterogeneity components, without blocking the possibility for areas in which more in-depth and more logically precise – but also more labor intensive ontology approaches – may become necessary. The discussion in the first section therefore suggests there is a common goal, but the approach of the Semantic Web may serve neither practically nor conceptually as the

exclusive basis of a digital library application such as vascoda. The following section demonstrates that this statement may be correct in the medium-term, but that newer developments in the Semantic Web suggest that the approaches may grow together over the longer term. Newer Semantic Web developments significant for the digital library Sharing ideologies vs sharing technology and standards Berners-Lee et al. (2001) articulate one of the most well known visions of Semantic Web, indicating that it should be ‘‘. . .an extension of the current [web], in which information is given well-defined meaning, better enabling computers and people to work in cooperation’’. Herrmann (2007) and Ankolekar (2007) no longer see these visions as binding; instead, they replace this with a series of practical goals: The Semantic Web gives .

common, interoperable standards for machine processable data and metadata markup;

.

data interoperability across knowledge sources, applications, communities, organizations;

.

architecture for interconnected vocabularies and communities;

.

reasoning about distributed web information . . . (Ankolekar, 2007, slide 8).

It is no longer about images such as the now famous ‘‘silver bullet’’[5], but about standardisation work and joint development of general tools which can be used to realise the various visions[6]. Web ontology language (OWL) vs simple knowledge organization system (SKOS) The Semantic Web community at W3C has become receptive to other theoretical approaches. This has been most evident in the use of ontology. Their differences with respect to the philosophy of thesauri and classifications have thus far represented a clear obstacle for digital library applications; however, there is a close connection with the Semantic Web community: that the deep and invisible web are important. Seen from a non-visionary standardisation perspective, the W3C ontologies are about the following: Idea: extend web markup standards to go beyond syntax and express structure and semantics of the data (a) ambitious goal: to enable intelligent software agents to reason about information on the web; (b)

less ambitious goal: to enable interoperability of data (Ankolekar, 2007, slide 6);

...

an ontology in computer science is essentially an engineering artifact . . .

(c) a formal and machine-executable model of a domain of interest; .

consists of a specific vocabulary to describe the domain of interest

.

a set of explicit assumptions regarding the intended meaning of the vocabulary in a logical language

(d) representing a shared understanding of the domain of interest; . to help humans and machines reach common understanding . can import information from other ontologies (Ankolekar, 2007, slide 6)

Semantic heterogeneity

243

LR 57,3

244

OWL Full, OWL DL, OWL Lite Since 2004, stable proposals have been established for ontology languages, fulfilling the requirement to support deduction processes. These are different in terms of potential deductive power (Ankolekar, 2007, slide 27). The varies levels of OWL expressiveness (‘‘Lite’’, ‘‘digital library (DL)’’, and ‘‘Full’’) exist as a result of the desire to have the necessary power of deductive languages, but to balance the resulting increased effort and difficulties of implementation (Figure 4). Language complexity and expressiveness therefore decreases from ‘‘Full’’ to ‘‘Lite’’: .

‘‘Full’’ is the whole thing,

.

‘‘description logic (DL)’’ restricts Full in some respects,

.

‘‘Lite’’ restricts DL even more (Herman, 2007b, slide 111).

Simultaneously, there was a desire not to compromise on the ability to exchange information between all types of ontologies on the Semantic Web. Simple knowledge organization system (SKOS)[7] The advocates of Semantic Web expected a broad cooperative development of ontologies through the introduction of OWL standards; however, such a cooperative, voluntary (and unpaid) joint endeavor has not, in fact, developed in ontologies over the last ten years. The idea underpinning the cooperative development of large information collections may indeed be successful on the Web; after all, Wikipedia is a fine example of this. However, writing partial ontologies which automatically connect (thanks to standardisation and tools) appears to be not such an interesting and stimulating activity for scientists. It can hardly be expected that this will change after ten years of intensive effort. This conclusion may be the main reason for Semantic Web advocates seeking alternatives. Ultimately, the desire for stronger deductive processes has been partially sacrificed via the development of the SKOS (Miles and Brickley, 2005). The only promising strategy therefore remains utilising the existing thesauri and classifications, even if they are not ‘‘true’’ ontologies (also Miles, 2006). This is where the Semantic Web meets the semantic heterogeneity components of the Shell model and the cross-concordances

Figure 4. The varying levels of OWL expressiveness

of the digital libraries. Integration of those is possible because the emphasis has shifted away from the previous visions and towards the standardisation and implementation of commonly used technologies. The explanations for ‘‘webifying’’ thesauri and classifications have nothing to do with the theoretical explanations for employing classifications and thesauri in information science (see the following section): However: ontologies are hard: .

a full ontology-based application is a very complex system

.

hard to implement, may be heavy to run. . .

.

. . . and not all applications may need it! (Herman, 2007b, slide 111)

OWL’s precision is not always necessary or even appropriate: .

OWL is a sledge hammer/SKOS a nutcracker’’, or ‘‘OWL a Harley/SKOS a bike

.

they complement each other, can be used in combination to optimise cost/benefit

Role of SKOS is: .

to bring the worlds of library classification and Web technology together

.

to be simple and undemanding enough in terms of cost and required expertise’’ (Herman, 2007b, slide 128)

Formal ontologies (like OWL) are important, but use them only when necessary: .

you can be a perfectly decent citizen of the Semantic Web if you do not use ontologies, not even RDFS. . . (Herman, 2007a, slide 51)

Languages should be a compromise between rich semantics for meaningful applications, feasibility, implementability (Herman, 2007b, slide 95)

SKOS-rationale vs information science rationale The information science explanation for using classifications or thesauri instead of the richer semantics of ontologies, is different from that cited above by the W3C–Group. As already mentioned, thesauri do without the richer semantic relations so that the effort needed for indexing remains small. The theory being that the additional gain in precision in individual cases does not justify the extra effort. The conceptual background is that in information retrieval, some semantic content remains unanalysed with regard to the search terms of the user and the descriptors used for indexing. Human intelligence supplements these non-analysed parts during machine interaction. The question in traditional information retrieval and in the Shell model is not whether parts of the semantic content remain unanalysed, but whether the (intelligent) user is able to supplement these parts with acceptable, minor effort. With SKOS, this can serve directly as a theoretical justification for the new W3C variant of ontologies. The newer developments of the Semantic Web, however, no longer requires acceptance of these theoretical thoughts as a necessary condition. SKOS can, but need not be, theoretically founded in a library and information science sense. In the Semantic Web, SKOS is currently accepted on a purely pragmatic basis as a way to achieve technical interoperability. In reality, there is hardly another possibility for the Semantic Web community than to permit this form of ‘‘weak’’ semantic foundation. One can therefore expect SKOS to rise in popularity.

Semantic heterogeneity

245

LR 57,3

246

Ontology sharing and statistical handling of vagueness Following the acceptance of SKOS by the Semantic Web community, there remains no standardisation activities within the Semantic Web that revolve around the quantitative-statistical, non-deductive vagueness components in information retrieval. At the International Conference on Semantic Web and Digital Libraries 2007 in Bangalore, Herman (2007b) reported that an incubation group had formed at the W3C. This group is working on alternative approaches based on description logic (fuzzy and statistical approaches). This development, combined with the existing Semantic Web portfolio, opens a path of enquiry which appears indispensable for digital libraries. .

Fuzzy logic — look[ing] at alternatives of description logic based on fuzzy logic — alternatively, extend[ing] RDF(S) with fuzzy notions

.

Probabilistic statements — hav[ing] an OWL class membership with a specific probability — combin[ing] reasoners with Bayesian networks (Herman, 2007b, slide 46)

Conclusion Recent developments in the Semantic Web mitigate many of the objections often raised by the digital library community against participation in its development. These recent developments also allow the Shell model to be reformulated, using these developments as a basis. The advantages are obvious. Whereas standardisation, for instance with the science portal vascoda, had to be negotiated among all participating partners in the group, all offers for integration which adhere to the W3C standards, and ideally use their general tools, would be accessible without entering into further time consuming negotiations. This line of development is currently not technologically refined for commercial applications, as initial applications of SKOS and associated tools have so far demonstrated. This is why it is not yet feasible to replace sowiport’s or the HTS with a SKOS model. Nevertheless, in the mid-term a reinterpretation of these approaches and an adaptation to the standards and tools supported by the W3C should be the best solution for the heterogeneity problem of digital libraries, like sowiport and vascoda. It is therefore recommended that all current developments should be checked for their adaptability to the W3C standards of the Semantic Web group. Ideally, test applications should be developed using the SKOS and SPARQL. In September 2007 the executive committee of vascoda (representing about 40 participants and providers of scientific information in Germany) decided to support the thesis ‘‘vascoda goes Semantic Web’’. This decision may prove advantageous for all participating information brokers, libraries and information scientists. They will benefit from the above described advantages of a reinterpretation of sowiport’s and vascoda’s heterogeneity approach. But it is ultimately beneficial to the Semantic Web, which will gain a committed digital library development group, as well as a large interconnected collection of 62 crosswalks containing a major quantity of the semantic knowledge in libraries and documentation centres in Germany. Notes 1. Generally, the shell model refers to the existence of various levels of indexing, quality and consistency into which documents may be grouped. The specific technique for

2. 3. 4. 5.

6. 7.

handling semantic heterogeneity between the different shells was developed as a refinement mainly since 2000. Today it is a central part of this model (Krause, 2006). SWD is the keyword-norm data file created through the cooperation of German scientific universal libraries on the basis of the RSWK (rules for the keyword catalogue). See Mayr et al. (2008) in this same journal for a detailed explanation of these added services. This thesis has to be seen in connection with the empirical finding that the user typically repeats multiple queries on the same search in succession. ‘‘. . . the next generation of the web, called the Semantic Web. To achieve even some of the promises for these technologies, we must develop vastly improved solutions for addressing the Grand Challenge of Information Technology, namely dealing better with semantics. . .. This challenge has been calling out for a silver bullet since the beginning of modern programming’’ (Fensel, 2001, p. V). The symbol for this perspective is the oft-cited ‘‘Semantic Web Layer Cake’’ (Ankolekar, 2007, slide 9). An overview of projects with SKOS can be found at: http://esw.w3.org/topic/SkosDev/ DataZone

References Ankolekar, A. (2007), Tutorial ‘‘Semantic Technologies for Digital Libraries’’, Semantic Web & Digital Libraries, Proceedings ICSD, International Conference, Bangalore, 21-23 February 2007. Berners-Lee, T., Hendler, J. and Lassila, O. (2001), ‘‘The semantic web’’, Scientific American, Vol. H. 5, pp. 34-43. Depping, R. (2007), ‘‘vascoda.de and the system of the German virtual subject libraries’’, Semantic Web & Digital Libraries, Proceedings ICSD, International Conference, Bangalore, pp. 304-14. Fensel, D. (2001), Ontologies – A Silver Bullet for Knowledge Management and Electronic Commerce, Springer, Berlin. Hahn, U. and Schulz, S. (2004), ‘‘Building a very large ontology from medical thesauri’’, in Staab, S. and Studer, R. (Eds), Handbook on Ontologies, Springer, Berlin, pp. 133-50. Herman, I. (2007a), ‘‘State of the Semantic Web’’, Key Note Speech Semantic Web & Digital Libraries, Proceedings ICDS, International Conference, Bangalore, 21-23 February 2007. Herman, I. (2007b), Tutorial ‘‘Introduction to the Semantic Web’’, Semantic Web & Digital Libraries, Proceedings ICDS, International Conference, Bangalore, 21-23 February 2007. Knorz, G. and Rein, B. (2005), ‘‘Semantische Suche in einer Hochschulontologie’’, Information, Wissenschaft & Praxis, Vol. 56 No. 5. Krause, J. and Stempfhuber, M. (2005), ‘‘Nutzerseitige Integration sozialwissenschaftlicher Textund Dateninformationen aus verteilten Quellen’’, in Arbeitskreis Deutscher Markt- und Sozialforschungsinstitute (ADM), Arbeitsgemeinschaft Sozialwissenschaftlicher Institute (ASI), Statistisches Bundesamt, Wiesbaden (Eds), Datenfusion und Datenintegration 6, Bonn, S. pp. 141-58. Krause, J. (2006), ‘‘Shell model, semantic web and web information retrieval’’, in Harms, I., Luckhardt, H.-D. and Giessen, H.W. (Eds), Information und Sprache, Beitra¨ge zu Informationswissenschaft, Computerlinguistik, Bibliothekswesen und verwandten Fa¨chern, Festschrift fu¨r Professor Dr. Harald H. Zimmermann, K.G. Saur, Mu¨nchen, S. pp. 95-106. Krause, J. (2007), ‘‘The concepts of semantic heterogeneity and ontology of the semantic web as a background of the German science portals vascoda and sowiport’’, Semantic Web & Digital Libraries, Proceedings ICDS, International Conference, Bangalore, 21-23 February 2007, pp. 13-24.

Semantic heterogeneity

247

LR 57,3

248

Kuhlen, R. (2004), Kapitel A 1: ‘‘Information’’, in Kuhlen, R., Seeger, T. and Strauch, D. (Eds), Grundlagen der praktischen Information und Dokumentation, Band 1, Handbuch zur Einfu¨hrung in die Informationswissenschaft und -praxis, Saur, Mu¨nchen, Ausgabe 5, pp. 3-20. Mayr, P. and Walter, A.-K. (2007a) ‘‘Zum Stand der Heterogenita¨tsbehandlung in vascoda: Bestandsaufnahme und Ausblick’’, in Bibliothek & Information Deutschland (Ed.), Information und Ethik 3. Leipziger Kongress fu¨r Information und Bibliothek, 19-22 Ma¨rz 2007, Verlag Dinges und Frick, Leipzig, available at: www.opus-bayern.de/bib-info/ volltexte/2007/290/ (accessed 19 December 2007). Mayr, P. and Walter, A.-K. (2007b), ‘‘Einsatzmo¨glichkeiten von Crosskonkordanzen’’, in Stempfhuber, M. (Ed.), Lokal-Global, Vernetzung wissenschaftlicher Infrastrukturen, 12. Kongress der IuK-Initiative der Wissenschaftlichen Fachgesellschaft in Deutschland, Tagungsberichte, GESIS-IZ Sozialwissenschaften, Bonn, pp. 149-66, available at: www.gesis.org/Information/Forschungsuebersichten/Tagungsberichte/Vernetzung/MayrWalter.pdf (accessed 19 December 2007). Mayr, P., Mutschke, P. and Petras, V. (2008), ‘‘Reducing semantic complexity in distributed digital libraries: treatment of term vagueness and document re-ranking’’, Library Review, Vol 57 No. 3. Miles, A. (2006), ‘‘SKOS – requirements for standardization’’, International Conference on Dublin Core and Metadata Applications, Mexico, 3-6 October 2006. Miles, A. and Brickley, D. (2005), ‘‘SKOS Core guide’’, W3C Working Draft, 2 November 2005 (Work in progress), available at: www.w3.org/TR/swbp-skos-core-guide/ (accessed 19 December 2007). Prasad, A.R.D. and Madalli, D.P. (Eds) (2007), ‘‘Semantic web & digital libraries’’, Proceedings ICDS, International Conference, Bangalore, 21-23 February 2007. Staab, S. and Studer, R. (2004), Handbook on Ontologies, Springer, Berlin. Stempfhuber, M. (2006), ‘‘Data integration in current research information systems’’, Magalha˜es, de S. T., Santos, L., Stempfhuber, M., Fugl, L. and Alrø, B. (Eds), CRIS-IR 2006, Proceedings of the International Workshop on Information Retrieval on Current Research Information Systems, Copenhagen, Denmark, 9 November 2006, Minho, pp. 29-51. Stuckenschmidt, H. and Harmelen, F. van (2005), Information Sharing on the Semantic Web, Springer, Berlin. Umsta¨tter, W. and Wagner-Do¨bler, R. (2005), Einfu¨hrung in die Katalogkunde - Vom Zettelkatalog zur Suchmaschine, Hiersemann, Stuttgart. Corresponding author Ju¨rgen Krause can be contacted at: [email protected]

To purchase reprints of this article please e-mail: [email protected] Or visit our web site for further details: www.emeraldinsight.com/reprints

Book reviews Ethics, Accountability and Recordkeeping in a Dangerous World Richard J. Cox Facet Publishing London 2006 298 pp. ISBN 978-1-85604-596-4 £44.95 Keywords Ethics, Records management, Archives, Professional ethics Review DOI 10.1108/00242530810865510

Book reviews

249

Ethics, Accountability and Recordkeeping in a Dangerous World by Richard J Cox is the first in a new series published by Facet called ‘‘Principles and Practice in Records Management and Archives’’. The series promises to be an interesting one, with titles Preserving Archives by Helen Forde and Management Skills for Archivists and Records Managers by Elizabeth Shepherd and Karen Anderson already announced. Given that Forde and Shepherd are such well-known names in the UK archives world, I am curious as to why Cox’s book was used to launch the series. Not to say that Cox is an unknown: Richard Cox is an eminent American academic specialising in the area of archives and records management and contributes regularly to specialist periodicals (primarily American) and is the author of numerous books. This meander through a range of threats facing archives and records management and its practitioners at the beginning of the twenty-first Century, a compilation of essays written between 2000 and 2005, makes interesting and, at times, uneasy reading. Despite its title, the book is really an exploration of this succinct statement which opens the book’s concluding chapter: The importance of records is both growing and being challenged in our society, as threats of terrorism, cynicism about truth, corporate greed, and faith in technology generate new forms of government secrecy, censorship and intellectual property controls.

Using this sentence as a starting point makes, to me, more sense than looking at the book’s title since Ethics, Accountability and Recordkeeping in a Dangerous World is misleading. ‘‘Danger’’ is first encountered 109 pages into the book, and ‘‘ethics’’ and ‘‘accountability’’ appear even further in. Indeed, even the reference to recordkeeping is not entirely accurate since the book looks at records in their fullest definition and context and not exclusively from a recordkeeping point of view. No matter. Despite this slightly disconcerting titling issue, the book is a fascinating look at each of the areas mentioned in the quote above and approaches each one in a different, thought-provoking chapter – the changing meaning of the term ‘‘information’’; the increasing use of technology in managing information; records, privacy, intelligence and terrorism; the changing role of the archivist and records manager; the meaning of the term ‘‘truth’’; censorship; and intellectual property control. The book finishes with a conclusion looking at the different meanings of the word ‘‘archive’’ by different constituencies, such as storage, data, memory, stories, old stuff, etc.

Library Review Vol. 57 No. 3, 2008 pp. 249-258 # Emerald Group Publishing Limited 0024-2535

LR 57,3

250

Stuck in the middle of these are two seemingly misplaced chapters which appear to be entirely the author’s own views on the US Presidential Library system and the method of appointment of the Archivist of the US. These chapters very clearly illustrate that, as I have said, rather than being one dedicated piece of work, the book is a collection of earlier essays. I personally found the discontinuity this caused a considerable obstacle, contrasting with the book’s easy-reading nature and agreeable style of writing, as the chapters do not connect seamlessly with each other. I would have preferred one long continuous piece of prose since I read the whole book traditionally, from cover to cover (albeit over a number of sessions). The benefit of discrete stand-alone chapters, however, is that rather than being obliged to read the whole book, each chapter can be read in isolation and out of order with no significant loss of context. In the main the book challenges the traditional, perhaps comfortable, role of the archivist and records manager and identifies areas of concern that we practitioners should either be addressing or at least keeping an eye on, alert and ready for future developments. It reminds us that we, possibly more than others, are at a crossroads and that it behoves our profession to address the challenges that we face in this new century. Some of these challenges indeed related to ethics and danger, such as those exemplified by the Enron/Arthur Andersen scandal and the September 11 attacks. Others are a little more esoteric, such as the future role of records and their custodians in a world where the simultaneous control and lack of control of intellectual property on the internet can blur issues of ownership, rights and powers relating to information. When I first received this book for review, I did question whether almost 300 pages were needed to look at ethics, accountability and recordkeeping in a dangerous world. Having read the book, I would say that only a fraction of this number were actually dedicated to these subjects. The rest of the book was, however, well used in addressing the other issues mentioned (truth, censorship, technology, etc.) which I suspect will continue to be sources of hand-wringing for archivists and records managers for some time to come. Matthew Stephenson University of Salford, Salford, UK

The Stuff of Thought: Language as a Window into Human Nature Steven Pinker Allen Lane (Penguin Books) London 2007 xi þ 499 pp. ISBN 978-0-713-99741-5 hardback £25.00 Keywords Language, Linguistics, Cognition, Mental models Review DOI 10.1108/00242530810865529 The way we think, and think about the way we think, are central to our understanding of meaning, and by extension human identity and human nature. We usually assume

that the ways in which thinking takes place, and meaning created and used, are embodied in language. By that token it seems easy to infer that language is the stuff of thought and a window into human nature. How all this takes place is the business of this new book from MIT language and cognition researcher Steven Pinker (once of MIT and now Harvard University). Readers will know of his earlier books, and Pinker himself suggests that The Stuff of Thought not only provides a sequel to earlier books on language and mind (The Language Instinct, 1994, and Words and Rules, 1999) but also to earlier books on human nature (How the Mind Works, 1997, and The Blank Slate, 2002). Anyone, then, keen to read the next book from Pinker – and these range from academic specialists to the general reader – will be excited at this newcomer. Anticipating bestseller sales, Allen Lane have published this hardback at a competitive price, and there is also a trade paperback too (ISBN 978-1-846-14050-1, for more go to www.penguin.com). It has simultaneously been published in the USA by Viking Penguin of New York. Pinker’s arguments are well-founded in extensive reading in the specialist literature of linguistics and cognition, psychology and decision-making, risk and uncertainty, consciousness and philosophy. He draws widely on books and articles for his evidence and, by that token, The Stuff of Thought will provide a distinctive window in a lot of current (and not so current) thinking about thinking, language, and human nature. In the best sense, it is a book that distills and digests and, like the others, popularises ideas usually hidden away in monographs and learned journals. He ties these up with popular debate and journalism, and lays out a stall likely to tease, amuse, fascinate, and infuriate by turns readers who come to it. Reading Pinker is never boring and always an education, although his idiosyncratic approach – demotic and sesquipedalian by turns, straight filtering of expert stuff with rough-and-ready chutzpah – makes for a very uneven book, and one which at times is simply far too long. It is a long book and so I hope that snapshots of how Pinker goes about his task will give readers of this review an idea of what I mean. The argument starts with 9/11 and Shakespeare and how we use words to make meaning. Words help us deal with reality, with emotions, with social relationships. How we learn language as children is still remarkable. Words mean so many things, often at once, change in context, derive from and shape conceptual structures, represent causal and intentional connections between events, can take innumerable erroneous forms, can be evasively passive, and reify concepts of space and time. Perhaps language, and perhaps concepts (some distinction is always useful), are innate: a point that leads Pinker to rehearse some of the most influential theories of language and thought – Fodor’s nativism (concepts are innate), the polysemy of pragmatics (language arises in context – a sad movie, a fast date, a white fella, from wordplay) and linguistic determinism (how they speak shapes how they think and live, giving Pinker the chance to discuss the Whorfian hypothesis). These ideas and approaches are highly relevant to any formal study of language and meaning, convince some or even most of the time, and are well worth grappling with as a newcomer to the field. They have not gone away for more specialist readers, either, and Pinker is right to put them on the slab. He supports his evidence with references to current and recent research, all of which can be checked and pursued by way of an extensive bibliography (and which needs a good academic library). Later chapters follow this up – by considering what Kant said about consciousness and how we think about space and time, why thinking involves categories (boys will be boys, que sera sera), how mental models use quantifiers (like more sauce and more

Book reviews

251

LR 57,3

252

pebbles, distinctions between continuous and discrete processes), and how understanding time entails a knowledge not just of temporality but also teleology. At times it gets quite clever – Pinker on counterfactuals, for example, plausibly postHumean in analysing human reasoning, controversially populist in wondering how reliably intuitive dilutions of rationalistic explanations can take us. I rather like his point that we try to make sense of events using intuitive forms of physics and find them cloudy in the moral sphere. The work of specialists like Jackendorff and Schank, Chomsky and Fodor, Spellman and Kripke lurk in the references, drawing on cognition, decision-making, representation and semantic structures. I was pleased to see Grice’s conversational structures in chapter 8 (on language games, where a case for a bit of Wittgenstein might also be made). Two sets of ideas seem to emerge about language and meaning – that we use language in combinations or structures (syntax? logical implicatures?), not a new idea by any means but one Pinker gets across in a lively way with plenty of examples, as we have seen so far; and that we use metaphor a great deal. Again this is not a new idea, as the work of Lakoff and Gentner and others has shown. Using metaphors or analogies for, and in our ways of thinking is, Pinker argues, distinctively human: metaphors appear in everyday and in scientific language, as we ask whether the world is a stage, when we come out of the fast lane, regard love as a journey, or speak about consuming electricity. Combine combinations and metaphors and we can build up the scripts and narratives (following Schank) that enable us to function as effective cognitive beings and social animals. Two further chapters discuss names (are names really meaning, said the philosopher) and swearing and the obscene (why we use them, why we shock and are shocked), and then on to games people play (a readable chapter on the ways in which language can be used in sarcastic and oblique ways: I was wondering if you could pass the salt, when a lady says no she means maybe, how authority and face come into play, mutual knowledge and tact). The review so far has identified several reasons why you simply should buy this book – Pinker is popular and readable, he distills lots of interesting insights into how and why we use language, he provides lots of examples, he writes wittily, and what he says about metaphors and social forms of language will make you feel more informed and even more socially adept. But surely there is a downside. The rhetorician Quintilian in Roman times was said to have had a word for every and any aspect of language – zeugma, hysteron proteron, litotes, epanorthosis, parataxis. Impressive stuff and great for the specialist who wants a meta-lexicon to define phenomena. There is more than a streak of Quintilian in Pinker as he struggles to explain content-locatives and double-object datives in a section on verb constructions, earnestly taxonomises causative and anti-causative alternations when examining microclasses of concepts like hit and cut and break and touch, explains how basic players in causal scenarios are agonists (antagonists exert forces on them, apparently), and explores how dysphemistic (the opposite of euphemistic) some swearwords are. The language used to examine language is a specialist one and perhaps this explains and excuses his tone. Even so it is a wordy and at times pompous book, for all its subtlety, even at times when the ideas are relatively simple. It goes on far too long and could have said everything it has to say – which, distilled down, is really interesting – in about 200 pages, not just for readers who already know the underlying ideas and theories, but especially for newcomers to the field who may drop the book into their laps from fatigue.

All that said – and this is the chocolate bar rather than the scorpion sting in the tail – The Stuff of Thought has the ability to impress and startle, offers an eclectic blend of quotable quotes from impressive and relevant experts, and is a book to dip into, like a chrestomathy: a kind of treasury, like Gringott’s Bank in the Harry Potter series, where unusual objects and ideas might be found, to play with, develop, impress your friends with, and – if you are a student with a genuine interest in language, meaning and thought, or if you teach in the field, a lot of unavoidably useful ideas abound. One day all six books should be distilled down in this way and then we shall be able to take stock of Pinker’s real contribution to the debate. He is right to warn us that, ‘‘left to our own devices, we are apt to backslide to our instinctive conceptual ways’’ (page 439): no danger of that for any reader who takes Pinker’s work seriously. Stuart Hannabuss Aberdeen Business School, Robert Gordon University, Aberdeen, UK

Ethics and Technology: Ethical Issues in an Age of Information and Communication Technology, 2nd edition Herman T Tavani John Wiley & Sons Inc. Chichester 2007 xxviii þ 396 pp. ISBN 10: 0-471-99803-6 ISBN 13: 978-0-471-99803-7 paperback £29.99 Keywords Ethics, Technology, Information technology, Information ethics, Professional ethics Review DOI 10.1108/00242530810865538 The first edition of this work established itself as a substantial source of ideas and interpretation in its field. Things move fast in cyberlaw and cyberethics, as Tavani (professor of philosophy of Rivier College in the USA) knows, and so it good now to have a second edition. It has been awaited eagerly and does not disappoint. Tavani has updated the book with recent court decisions like Verizon v RIAA and MGM v Grokster, cases with influence far wider than merely in the USA itself. He also draws out current issues like privacy and surveillance, intellectual property rights (IPR) and open source software, bioinformatics and the legal and ethical implications of genomics research, and online pornography and young people. These are all mainstream ethical issues for practitioners and students/lecturers on relevant courses, and Tavani’s discussion of the issues (which includes scenarios) make his approach particularly thought-provoking and informative. With all this, you get a lot for your money in what is a sturdily-bound softback textbook.

Book reviews

253

LR 57,3

254

The shape of the book gives an idea of its scope and approach – cyberethics and how to look at/for them, a bit of ethical background and critical thinking, professional ethics and codes of conduct, cyberspace and privacy and security and cybercrime, intellectual property issues (including digital rights management (DRM)), commerce and free speech in cyberspace, the digital divide and work, identity and community, and technology-driven issues. This gives you an idea of how such a book will be of interest not just to readers like those noted above but also to students of computing and technology, the Internet, the media and journalism, sociology and criminology. Seven appendices are available online and include codes of ethics and other resources. Advice is also given to students and instructors. There is a glossary of key terms and the book is well indexed. Typical chapters provide definitions and cases (the first one is that of Verizon and the Recording Industry Association of America in Washington in 2003, and further scenarios include ones on software piracy and theft), perspectives (ways of looking at cyberethics – professionally, methodologically, philosophically), chapter summaries, review and discussion questions, and further readings. He sorts out ethics and morality and where they crossover, acknowledging that many people never get beyond relativism and ‘‘who am I to judge others?’’. Consequence, duty, contract and virtue theories get a quick but useful mention. How to think and argue critically and logically comes in chapter three. The chapter on professional codes of ethics and accountability is one that should not be side-stepped by any pre- and post-experience training course, and adds to the already substantial stuff out there from bodies like the American Library Association and others. I am glad to see them criticised, too, because codes of ethics are not selfevidently good things like motherhood and apple pie but deeply challenging and complex fields of engagement. The US context here does not constrain Tavani’s debate of the issues having wider resonance. Cases about whistle-blowing rightly move on to issues of professional liability, an area of law close to codes to ethics. Privacy is another unavoidable issue in such a book and Tavani acquits himself well here, picking up on the information theme but looking wider at cookies, radio frequency identification and data warehousing, likely to be of interest to readers in related fields. Such a wide-ranging topic as privacy is hard to manage concisely, and Tavani defines it clearly, asks why it is important above all today when new technology makes data monitoring and surveillance easier. Some interesting and plausible cases are presented for study, and the discussion includes data mining, search engines, and the effectiveness of privacy-protection tools. Privacy law is there but for Tavani, in an ethical context, the law is only one element in the mix. Again, extensive further reading is provided. Tavani picks out ethically relevant issues from what is a wealth of material on the technological and legal aspects of security (for instance, the ethics of hacktivism, the effectiveness of encryption, and whether we can expect total security in cyberspace). Under cybercrime come hackers and worms, the law and various scenarios that open up ethical choices for students, practitioners and companies. Identity theft (phishing and pharming) is where cybercrime and privacy converge. Corporate espionage and sting/entrapment are two more interesting issues. Surveillance (by Carnivore and other methods) throws up the challenges and controversies associated with the Patriot Act 2001 (about anti-terrorist interception): Tavani moves quickly on to international issues and biometrics.

One of the most important and complex dimensions of cyberethics, and information ethics generally, is the overlap between law and ethics. Time and time again this overlap provokes practitioners to wonder what is best in any situation. IPR and copyright law play an important part in software and digital music distribution, as cases like Grokster and DeCSS (music copyright and DVD decompilation, respectively) show. This could easily turn an ethics textbook into a law textbook, but Tavani takes care to avoid this – by examining the utilitarian theory of property (including IPR) for instance, by asking whether one’s personality is property, and by discussing Gnu’s Not Unix and the open source movement. The whole debate about copyright as a creative commons, in cyberspace and elsewhere, is as much ethical and social as legal. With an intellectual commons, might the ‘‘public domain’’ disappear, he asks. Lessig understandable hovers (in books like The Future of Ideas: The Fate of the Commons in a Connected World, New York, Random House, 2002). Final chapters deal with domain names and (as?) trademarks, online pornography and defamation, the liability of internet service providers, and the extent to which DRM might be used to enforce copyright rules. Interesting cases are provided, interestingly from sources that include Wikipedia (spelt incorrectly on page 265). Metatags, deep linking, spam, pornography – the usual suspects but ones that open up community standards as well as the law. There is, too, the wider issue of free speech and how far it can and should go. Understandably for information readers, hate speech, filtering, and liability are not far behind. Widening out globally, the digital divide presents social, political, economic and ethical challenges, and Tavani will provoke reaction with his remarks about race and gender and disability, employee monitoring in the workplace, the pluses and minuses of online or virtual communities, and virtual reality. In these ways, his book extends beyond traditional information readers to a much wider constituency, such as people interested in media, e-government, virtual reality, and remote working. In creating new opportunities to create avatars and new identities, and in posing new challenges of social fragmentation, these ‘‘developments’’ deserve full ethical investigation. It may even be that biochemical and genomic projects may not only redefine our sense of self but change community for the worse. This second edition, then, is wide-ranging, sympathetic to many different readers’ interests, a good working textbook for academic use, and on the button in making readers think hard about the issues. Tavani challenges us not only to examine the issues but to avoid complacency about the law dealing with things: it is only one way in which solutions might be found, and, like technology, is has a Faustian dimension – at times decisions are misguided and aggravate problems. It is all too easy to get heavy in books about ethics, particularly in the information profession where wanting to be right makes many of us earnest. Tavani’s book is worth buying in multiple copies where ethics and technology, and information ethics, are taught seriously, although instructors in countries other than the USA should expect to provide appropriate jurisdictional law and cultural ethical approaches to the issues. Stuart Hannabuss Aberdeen Business School, Robert Gordon University, Aberdeen, UK

Book reviews

255

LR 57,3

256

Computers in Libraries: An Introduction for Library Technicians Kate Wilson The Haworth Information Press Binghamton NY and London 2006 194 pp. ISBN 10: 0-7890-2151-X ISBN 13: 978-0-7890-2151-9 paperback US$19.95 Keywords Computers, Library Technicians, Information Technology Review DOI 10.1108/00242530810865547 Aimed at newly qualified as well as practising library technicians, this book provides a very thorough, if very basic, overview of technology and its use in libraries of the twenty-first century. Kate Wilson states that it is her aim to look at the way in which roles of library staff have changed in the last 20 years through the impact of computer technology, most notably the spread of the Internet. Looking in detail at library management systems, OPAC’s, resource sharing, information searching and skill-sharing, Wilson has provided a very succinct, approachable and readable text that should provide students and those staff new to the library with an indepth and thorough understanding of the way in which a library operates through its management systems. However, her claim that existing staff may wish to use the book in order to update their skills is perhaps overly enthusiastic for the topics covered, and the detail provided is very basic, despite its thoroughness. The opening chapter, introducing computers, gave me the impression that I was about to read a book that had been published 20 years ago, not one that was looking at 20 years of developments. Even so, the style of writing and the content are basic and easy to understand whilst not condescending. There are no expectations of prior knowledge which is very helpful. Despite this approach, however, the review questions found at the end of each chapter are rather demeaning and would be better placed in a child’s textbook than a book for mature adult library staff. For example the chapter on computers asks the reader to name five components of computer hardware. Later chapters are, however, an improvement; after discussing computers, Wilson moves on to look at the Internet and its application in the library. A general chapter introducing library management systems is followed by a chapter on each of the main aspects of a system – acquisitions, cataloguing, circulation and serials. There then follows chapters on OPAC’s; resource sharing (i.e. inter-library loan procedures); information searching – the use of databases both online and on CD/DVD; computer skills, and finally, the future for technology in library services. With the exception of some very basic and perhaps unnecessary advice about using the mouse and keyboard in the computer skills chapter, the content is useful in parts giving a good overview of the practical, behind-the-scenes work that librarians undertake. These aspects of the book would perhaps be a good introduction for first year Information Studies students or very new library technicians.

The glossary of terms that each chapter begins with is a useful quick reference, particularly when considering the use of Internet and online databases as there is a considerable amount of technical terminology and a number of acronyms used for describing this. The few diagrams and screen shots provided throughout the chapters provide a useful visual guide to some of the processes described in the text, the extensive bibliography and full index also allow for further reading and quick reference. Overall it is a very succinct book and each chapter is clearly divided allowing the reader to locate the information they require quickly. In summary this book contains some useful elements but is generally too basic to be recommended as a musthave for all library technicians. Louise Ellis-Barrett Downsend School, Leatherhead, UK

Blogging and RSS: A Librarian’s Guide Michael P. Sauers Information Today Medford, NJ 2006 272 pp. ISBN 978-1-57387-268-3 paperback US$29.50 Keywords Blogs, Libraries 2.0, Library technology, Internet Review DOI 10.1108/00242530810865556 Blogs in libraries are now a hot topic. Beginning with a very clear description of blogs, their types and their effect on traditional media and search engines, the author then explains why this important information resource and tool is an excellent means by which a library can fulfil its role as information provider. Calendars of events, news of recent acquisitions and other library news can be efficiently disseminated through a library blog. There are two chapters on the library blogosphere which ‘‘set the scene’’ for stimulating the reader to create his/her own blog. Out of hundreds of blogs written by librarians, library staff, library school students and other bloggers in the library and information field, the author has selected some of the best written, most informative and the funniest. Clear screen shots are provided of 34 of these blogs. The second chapter on the blogosphere includes interviews with the bloggers. People answer questions about; where they live, family life, their jobs, why they blog, reasons for starting blogging combined with a description of their blog. The bloggers come from different library sectors, including academic, special and public libraries as well as related areas such as Research and Development. What is most informative from these different perspectives are the comments about views on the strengths and problems with blogs as well as their top five favourite blogs to read. The most useful chapter for me is ‘‘Creating a Blog’’, with 45 screen shots. Different methods for creating a blog are explained, including important detailed information about blogger.com, a free resource. I would never have created my blog

Book reviews

257

LR 57,3

258

(librarianjourney.blogspot.com) without this clearly explained chapter. I share the author’s recommendation, that having read through the book once it is useful to then focus on specific chapters of relevance. In fact, it is a guide to be consulted regularly on blogs and RSS feeds. The two chapters on RSS feeds are slightly more demanding than the introduction to blogging. At least one public library has made its online subject guides available as feeds. There are also details of miscellaneous feeds and services which could be helpful for difficult reference enquiries or personal interest. With 41 clearly printed screen shots, there are plenty of tempting feeds to consider, though it may be necessary to take a few days off work to really indulge oneself in these temptations. The author has achieved his aim of providing practical advice and explaining how to get started at little or no cost. The text is supported by recommended reading of online and print articles, an appendix of feed code examples and a glossary of all the relevant ‘‘techie’’ jargon used by bloggers. This is an excellent reference tool for librarians who cannot afford to ignore blogs and blogging in information services, and it has certainly inspired me to embrace great technological innovations. The stress normally associated with tedious hours of trial and error techniques will not befall the person that wields a practical handbook such as this one. Margot Lindsay London Centre for Dementia Care, University College London, London, UK