ContactFinder: Extracting indications of expertise and answering questions with referrals

This paper presents a novel approach to utilizing large heterogeneous information repositories such as the World Wide Web, Lotus NotesTM, or Usenet. Rather than extracting knowledge to be used directly in problem solving, our approach is to extract key contacts for specific technical areas that can then be contacted for help in those areas. We discuss this in the context of ContactFinder, an intelligent agent that extracts key contacts and answers discussion questions with referrals. scheduling [Dent et. al., 1992; Maes and Kozierok, 1993; Kautz et. al., 1994], Usenet message filtering [Sheth, 1994], or other information search and retrieval domains [Holte and Drummond, 1994; Knoblock and Arens, 1994; Levy et. aL, 1994]. Like these other systems, ContactFinder extracts information from a large number of documents in order to present it to users in a more focused and productive fashion. Unlike these previous approaches, however, our goal is not to present the user with a subset of the information that can be used directly in problem solving. ContactFinder instead extracts key human contacts for different topic areas, and suggests contacts that can help users solve problems that arise.