OBIWAN - a visual interface for prompted query refinement

Abstract

Typically, users submit very simple search queries to digital document data collections. Often these queries can result in extremely broad answers or answers in which document relevance is hard to assess. Our group has developed a suite of tools which build sophisticated indexes to document collections. These tools can be used to provide cues to help users formulate more effective queries. In order to present these cues and manage the process of query refinement, we have developed a Java-based client server system which uses these indexes and tools. One of the most powerful parts of our system is its ability to recognize domain-specific multi-word names and terms, using heuristic methods, and to index these vocabulary items in such a way that we can look up vocabulary items that commonly are related to the original query terms. This system responds to an initial query by suggesting additional items that the user can use to focus the query.