Concept-based Interactive Query Expansion Support Tool (CIQUEST)

This report describes a three-year project (2000-03) undertaken in the Information Studies Department at The University of Sheffield and funded by Resource, The Council for Museums, Archives and Libraries. The overall aim of the research was to provide user support for query formulation and reformulation in searching large-scale textual resources including those of the World Wide Web. More specifically the objectives were: to investigate and evaluate methods for the automatic generation and organisation of concepts derived from retrieved document sets, based on statistical methods for term weighting; and to conduct user-based evaluations on the understanding, presentation and retrieval effectiveness of concept structures in selecting candidate terms for interactive query expansion. The TREC test collection formed the basis for the seven evaluative experiments conducted in the course of the project. These formed four distinct phases in the project plan. In the first phase, a series of experiments was conducted to investigate further techniques for concept derivation and hierarchical organisation and structure. The second phase was concerned with user-based validation of the concept structures. Results of phases 1 and 2 informed on the design of the test system and the user interface was developed in phase 3. The final phase entailed a user-based summative evaluation of the CiQuest system. The main findings demonstrate that concept hierarchies can effectively be generated from sets of retrieved documents and displayed to searchers in a meaningful way. The approach provides the searcher with an overview of the contents of the retrieved documents, which in turn facilitates the viewing of documents and selection of the most relevant ones. Concept hierarchies are a good source of terms for query expansion and can improve precision. The extraction of descriptive phrases as an alternative source of terms was also effective. With respect to presentation, cascading menus were easy to browse for selecting terms and for viewing documents. In conclusion the project dissemination programme and future work are outlined