Meta-Information for Knowledge Navigation and Retrieval: What's In There

Knowledge navigation in the business world is oftell a difficult and expensive process. Commercial databases are notoriously incomplete and inaccurate, access charges can run into hundreds of dollars per hour, and the client always wants the answer yesterday. In coping with such problems, professionals who search for information on a daily basis rely on metainformation that goes beyond standard descriptions of the content of data sources. For example, in working with information specialists at Price Waterhouse LLP (PW), a large professional services provider, one point we were advised, "If you’re looking for quick information about foreign companies that’s not too out-of-date, use database so-and-so because the vendor is offering free connect time this month." Such a heuristic does involve the source’s semantic content, of course, but also refers to resources (money) available for the retrieval, and quality (timeliness) of the retrieved data. If the tools we build are to make knowledge navigation in the business world any easier, they will have to exploit, or at least represent, this type of meta-information in choosing information search and retrieval strategies. A number of researchers have noted the importance of meta-information. KQML [Finin et al., 1994] provides an interprocess communication language for information agents, but makes few commitments about the semantics of the meta-information to be communicated. MCC’s Carnot project [Collet el al., 1991] introduced some of the necessary computational infrastructure with concept matching and resource models. The SIMS project [Arens el al., 1993] exploits some semantic and run-time recta-information to produce efficient query plans. Yet there has been little reported in the literature on exactly what meta-information is important for describing real-world task contexts. We present our preliminary findings on this subject in this paper. Our goal is to create a "content theory" on which to base languages and support tools for the delivery of knowledge services, in contrast to focusing more narrowly on answering queries in isolation. We define a knowledge service as the provision of content in some form and at some level of quality, subject to constraints on the resources needed to perform the service. This paper begins with some conceptual preliminaries, and then discuss these four elements in more detail. As illustrative examples of the use of meta-information, the paper describes two knowledge services: monitoring management changes, and search for potential conflicts of interest in prospective litigation support clients.