Dialogue for web search utilizing automatically acquired domain knowledge

Abstract

Web search has become an everyday activity for millions of people. Unfortunately, there is a number of well known problems associated with it. Two of those problems are to express an information need as a set of search terms and to actually present only the most relevant matches for this query once the search has been performed. This paper tackles those problems for limited domains where the amount of data might be large, but the application of some fairly simple intelligent techniques to extract conceptual information is feasible. We describe how a dialogue system utilizes the extracted knowledge to guide the search. This paper builds on earlier work which focussed on the extraction of knowledge-rich indices.