A Computational Lexicon Based on Intrinsic Nature of Word Senses: A Bilingual Approach

The objective of the proposed project is to explore an alternative classification of word senses based on the intrinsic nature of words, so as to complement the conventional conceptual classifications of words and to enrich the information in computational lexicons for use in a variety of natural language processing tasks. The project will take a bilingual approach, starting with a detailed investigation of polysemy of frequent nouns and verbs in English and Chinese. Their intrinsic nature will be studied from both a psycholinguistic perspective, such as via word association tests, and a computational perspective, such as experimenting with various lexical information for automatic word sense disambiguation (WSD). A customisable WSD system will be developed, for performing experiments with various lexical information types, and for verifying the usefulness of the resulting lexicon, with intrinsic properties of words captured, by getting feedback from it to fine-tune the combination of information in further WSD tasks.