Context expansion with Global keywords for a Conceptual Density-based WSD

Abstract

The resolution of the lexical ambiguity, which is commonly referred to as Word Sense Disambiguation, is still an open problem in the field of Natural Language Processing. An approach to Word Sense Disambiguation based on Conceptual Density, a measure of the correlation between concepts, obtained good results with small context windows. This paper presents a method to integrate global knowledge, expressed as global keywords, in this approach. Global keywords are extracted from documents using a model based on term frequency and distribution. Preliminary results show that a slight improvement in recall can be obtained over the base system. © Springer-Verlag Berlin Heidelberg 2005.