Adapting A Diagnostic Problem-Solving Model To Information Retrieval

In this paper, a competition-based connectionist model for diagnostic problem-solving is adapted to information retrieval. In this model, we treat documents as `disorders\u27 and user information needs as `manifestations\u27, and a competitive activation mechanism is used which converges to a set of documents that best explain the given user information needs. By combining the ideas of Bayesian inferencing and diagnostic inferencing using parsimonious covering theory, this model removes many difficulties of direct application of Bayesian inference, such as the unrealistically large number of conditional probabilities required in the knowledge base, the computational complexity, and certain unreasonable independence assumptions. Also, Bayesian inference strengthens the parsimonious covering theory by providing a likelihood measure which can be used to rank documents as well as to guide the retrieval to the most likely set of documents. We also incorporate thesaurus information to provide semantic relevance among the index terms. Our experimental results using 4 standard document collections demonstrate the efficiency and the retrieval effectiveness of the thesaurus-based model, comparable to or better than that of various information retrieval models reported in the literature