Integrating conceptual density with wordnet domains and CALD glosses for noun sense disambiguation

Abstract

The lack of large, semantically annotated corpora is one of the main drawbacks of Word Sense Disambiguation systems. Unsuper-vised systems do not need such corpora and rely on the information of the WordNet ontology. In order to improve their performance, the use of other lexical resources need to be investigated. This paper describes the effort to integrate the Conceptual Density approach with sources of lexical information different from WordNet, particularly the WordNet Domains and the Cambridge Advanced Learner's Dictionary. Unfortunately, enriching WordNet glosses with samples of another lexical resource did not provide the expected results.