CWI at INEX 2002

This paper describes our participation in INEX 2002 (the XML Retrieval Initiative) and discusses several aspects of our XML retrieval system: the retrieval model, the document indexing and manipulation scheme and our preliminary evaluation results of the submitted three runs. In our system, we have used a probabilistic retrieval model where we map (structural) properties of documents to dimensions of relevance and use these dimensions of relevance for retrieval purposes. The study concentrates on coverage, defined as the amount of relevant information present in a document component. We also discuss an efficient and database-independent indexing scheme for XML documents, based on text regions and discuss region operators for selection and manipulation of XML document regions. Initial evaluation of our results, with a rather adhoc approach, made clear that evaluation measures for structured document retrieval needs more discussion and research.