An Ordering of Terms Based on Semantic Relatedness

Abstract

Term selection methods typically employ a statistical measure to filter or weight terms. Term expansion for IR may also depend on statistics, or use some other, non-metric method based on a lexical resource. At the same time, a wide range of semantic similarity measures have been developed to support natural language processing tasks such as word sense disambiguation. This paper combines the two approaches and proposes an algorithm that provides a semantic order of terms based on a semantic relatedness measure. This semantic order can be exploited by term weighting and term expansion methods.