An Iterative Approach to Estimating Frequencies over a Semantic Hierarchy

This paper is concerned with using a semant ic hierarchy to es t imate the frequency with which a word sense appears as a given argument of a verb, assuming the da ta is not sense disambiguated. The s tandard approach is to split the count for any noun appearing in the da t a equally among the alternat ive senses of the noun. This can lead to inaccurate estimates. We describe a rees t imation process which uses the accumulated counts of hypernyms of the al ternat ive senses in order to redis t r ibute the count. In order to choose a hypernym for each alternative sense, we employ a novel technique which uses a X 2 test to measure the homogeneity of sets of concepts in the hierarchy.