Word Sense Disambiguation Using WordNet Relations

In this paper, the “Weighted Overlapping” Disambiguation method is presented and evaluated. This method extends the Lesk’s approach to disambiguate a specific word appearing in a context (usually a sentence). Sense’s definitions of the specific word, “Synset” definitions, the “Hypernymy” relation, and definitions of the context features (words in the same sentence) are retrieved from the WordNet database and used as an input of our Disambiguation algorithm. More precisely, for each sense of the word a sense bag is formed using the WordNet definition and the definitions of all the “Hypernyms” associated with the nouns and verbs in the sense’s definition. A similar technique is used, for all the context words and the definitions of the “Hypernyms” (associated with the context nouns and verbs), to form a context bag. Then, a technique of assigning weights to words is applied. The weight for every word is inversely proportional to the hierarchy depth in the WordNet taxonomy of the associated “synset”. Eventually, the disambiguation of a word in a context is based on the calculation of the similarity between the words of the sense bags and the context bag. The proposed method is evaluated in disambiguating all the nouns for all the sentences in the Brown files.