Report on ECDL 2003

Report on the 7th European Conference on Research and Advanced Technology for Digital Libraries (ECDL 2003), Trondheim, Norway, 17-22 August 2003

Michael Day, UKOLN, University of Bath.

The Seventh European Conference on Research and Advanced Technology for Digital Libraries (EDCL 2003) was held on the 17-22 August 2003 at the Britannia Hotel in Trondheim, Norway. This review will attempt to give a flavour of the event; describing the invited speaker presentations together with some highlights from the rest of the conference.

Each of the three days of the main conference commenced with a lecture by an invited speaker. These were by John Lervik of the search company FAST (Fast Search & Transfer), Clifford Lynch of the Coalition for Networked Information, and Karen Sparck Jones, Emeritus Professor at the University of Cambridge. After some initial comments on the differences between search engines and databases, Lervik looked at the prospects for future developments in search engine technology. This included a look at the new relevance model used by FAST for its third generation search engines, which uses algorithmic techniques like automatic entity extraction, structural document analysis, linguistic query analysis, etc., for improving both the precision and recall of searches.

Lynch's talk was entitled 'Stewardship in the digital age,' and he provided some characteristically thoughtful reflections some of the issues that surround the long-term preservation of digital information. Before going on to consider digital stewardship in more detail, he reminded delegates that the physical world remained fragile, noting the continuing risks to cultural memory institutions from war, fire and flood, and other disasters. One difference, however, is that we now have the means of creating digital surrogates of such content and replicating it. Replication and the geographical distribution of storage also form a key part of Lynch's proposed approach to digital preservation. This would, however, only represent the first step; Lynch arguing that curation itself could also be distributed - thus reducing the risks from bad decision-making. Lynch finished his talk with some comments on how these principles could be applied by nation states and how the growing amount of digital information in personal collections (e-mail, digital photographs, etc.) would begin to change assumptions about these problems.

Professor Sparck Jones reviewed progress in information retrieval (IR) research and looked at its implications for digital library development. The opening part of her talk was mainly historical, looking back at the development of IR research. She argued that by the 1960s, researchers had demonstrated the theoretical value of using statistical methods for text retrieval, but that such techniques had been implemented in few large-scale practical applications until the further development of network technologies in the 1990s. For example, IR research had some influence on the development of Web search services like AltaVista. In the meantime, however, operational systems remained largely focused on controlled language indexing and boolean matching. Sparck Jones recommended that those developing digital library systems should begin to take over and integrate statistical methods, e.g. for retrieval, speech recognition, summarising, and question answering.

The other sessions of the conference mainly comprised panel discussions and presentations of the peer-reviewed papers. Perhaps the liveliest - if ultimately fruitless - panel discussion was the first on the future of academic publishing, where Leo Waaijers of Wageningen University and David Prosser of SPARC Europe took on Michael Mabe of Elsevier Science in a two-pronged attack on the subscription-based scholarly publishing model. Prosser argued that the current publishing model was not working efficiently, either for publishers or the authors and readers of papers, and proposed the adoption of the 'open-access' publication model, paid for by author charges. Waaijers attempted to use mathematics to show how the price per article was much higher for commercially published journals, e.g. when compared with the author charges levied by the open-access publisher BioMed Central. He also argued that licensing deals had very little effect on these prices. Mabe countered that generous license deals meant that not all subscribers paid the full list price of journals, and noted that there were currently about 10 million users of ScienceDirect. He also stressed that commercial publishers invested heavily in service reliability, e.g. to guarantee 100% availability, and that Elsevier gave free-access to ScienceDirect to scholars in developing countries.

The second panel discussion was chaired by Stefan Gradmann of the University of Hamburg and concerned the use of the term 'digital libraries' as a concept or metaphor. Gradmann argued that the use of language was important, especially in political contexts, and wondered whether the digital library metaphor was counter-productive or even dangerous. At the end of the discussion, however, I'm not sure if anyone had come up with a better term, and one contributor commented that there was some value in a consistent 'brand.' Andreas Rauber of Vienna University of Technology chaired a third panel session on the importance of metadata for digital preservation. All panellists seemed to agree that metadata was crucial, but that there now needed to be a stronger emphasis on implementation and for the experience gained thereby to be fed back into the development of standards.

Over 50 posters and demonstrations had also been accepted for presentation at the conference. In the last session of the first day of the workshop, the authors of these were given one minute to introduce them in a plenary session - which seemed to be a successful way of drawing the conference participants' attention to a range of interesting work in progress.

In his welcome message to the conference, programme chair, Traugott Koch of NetLab, Lund University Libraries, noted that out of 161 paper submissions, 47 had been accepted for presentation at the conference - an acceptance rate of just over 29%. The paper sessions were divided into two parallel tracks. Interesting topics included the consistency and quality of data in metadata harvesting contexts, the use of automatic software tools, e.g. for text categorisation and topical Web crawling, and the preservation challenges of new media artworks. Springer-Verlag has published the peer-reviewed papers in its series Lecture Notes in Computer Science (LNCS 2769), while presentation slides from the invited talks and many of the papers, and details of panels, posters, tutorials, and workshops are available from the conference Web site (http://www.ecdl2003.org/).

As is usual with ECDL, there was no shortage of events going on in conjunction with the conference. Before the conference started, there were eight tutorials on a diverse range of topics, including thesauri and ontologies, music retrieval, and the CIDOC Conceptual Reference Model. On the days after the conference, five workshops were held, subjects including networked knowledge organisation systems, digital libraries in healthcare, and Web archives. In addition, there was evidence of many other informal meetings going on in parallel with the conference.

The social events were also successful; a reception at the historic Archbishop's Palace to mark the opening of the conference, another at the Britannia Hotel to accompany the poster and demonstration session, and an excellent conference dinner in Palmehaven with entertainment provided by students from the Norwegian University of Science and Technology. In addition, there was some spare time for delegates to take in some of the sights of Trondheim; the Nidaros Cathedral, the Trøndelag Folk Museum at Sverresborg, and the short boat trip to Munkholmen in Trondheim Fjord.

The next ECDL will be held on the 12-17 September 2004, in Bath, United Kingdom (http://www.ecdl2004.org/).