Estimating Probability Distributions over Hypotheses with Variable Unification

Abstract

We analyze the difficulties in applying Bayesian belief networks to language interpretation domains, which typically involve many unification hypotheses that posit variable bindings. As an alternative, we observe that the structure of the underlying hypothesis space permits an approximate encoding of the joint distribution based on marginal rather than conditional probabilities. This suggests an implicit binding approach that circumvents the problems with explicit unification hypotheses, while still allowing hypotheses with alternative unifications to interact probabilistically. The proposed method accepts arbitrary subsets of hypotheses and marginal probability constraints, is robust, and is readily incorporated into standard unification-based and frame-based models. 1 Introduction The application of Bayesian belief networks (Pearl 1988) to natural language disambiguation problems has recently generated some interest (Goldman & Charniak 1990; Charniak & Goldman 1988, 1...