Fuzzy data mining of term associations for flexible query answering.
Henrik Legind Larsen · 2001
We present an approach to data mine a document base for term associations that are useful for query expansion in flexible query answering. Fuzzy logic and probability theory are applied to handle key aspects of uncertainty arising from vagueness of terms and randomness of term occurrence. In particular, we consider conditional probability to increase the utility of association, Chebyshev's inequality to reflect statistical confidence, and the relative cardinality measure, introduced by Delgado et al., to handle the vague concept of a set of documents characterized by a particular term.