Report on the INEX 2003 Workshop, Schloss Dagstuhl, 15-17 December 2003

摘要

The INEX workshop showed that XML retrieval is a challenging new field within IR research. The partici-V pating groups are applying a broad range of approaches, most of which are extensions of models developed for "flat" text documents. Since these extensions often include a variety of new tuning parameters, the retrieval results presented so far do not yet allow for judging about the quality/suitability of the different methods. In addition to learning more about XML retrieval approaches, the workshop allowed participants to contribute to the evaluation methodologies. This has resulted in a number of participating groups to not only continue to develop their XML approaches, but to be actively involved in various aspects of the evaluation tasks (i.e topic format syntax + parser, metrics, etc). INEX 2004 will start in March of this year, and in addition to the standard ad-hoc task, has 4 new tracks: 1. interactive track focusing on interactive XML retrieval, considering also navigation through the hierarchical structure, 2. heterogeneous collection track, comprising various XML collections from different digital libraries, as well as material from other computer science-related resources, 3. relevance feedback track dealing with relevance feedback methods for XML, 4. natural language track where natural language formulations of CAS queries have to be answered. In the future, we are planning to look at XML-based multimedia test-beds, as this is an important issue in many applications such as in those in the context of digital libraries.