Query Refinement by Relevance Feedback in an XML Retrieval System

Abstract

In recent years, ranked retrieval systems for heterogeneous XML data with both structural search conditions and keyword conditions have been developed for digital libraries, federations of scientific data repositories, and hopefully portions of the ultimate Web. These systems, such as XXL [2], are based on pre-defined similarity measures for atomic conditions (using index structures on contents, paths and ontological relationships) and then use rank aggregation techniques to produce ranked result lists. An ontology can play a positive role for term expansion [2], by improving the average precision and recall in the INEX 2003 benchmark [3].

Due to the users’ lack of information on the structure and terminology of the underlying diverse data sources, and the complexity of the (powerful) query language, users can often not avoid posing overly broad or overly narrow initial queries, thus getting either too many or too few results. For the user, it is more appropriate and easier to provide relevance judgments on the best results of an initial query execution, and then refine the query, either interactively or automatically by the system. This calls for applying relevance feedback technology in the new area of XML retrieval [1].

The key question is how to appropriately generate a refined query based on a user’s feedback in order to obtain more relevant results among the top-k result list. Our demonstration will show an approach for extracting user information needs by relevance feedback, maintaining more intelligent personal ontologies, clarifying uncertainties, re-weighting atomic conditions, expanding query, and automatically generating a refined query for the XML retrieval system XXL.