CITYU-HIF: WSD with Human-Informed Feature Preference

This paper describes our word sense disambiguation (WSD) system participating in the SemEval-2007 tasks. The core system is a fully supervised system based on a Naïve Bayes classifier using multiple knowledge sources. Toward a larger goal of incorporating the intrinsic nature of individual target words in disambiguation, thus introducing a cognitive element in automatic WSD, we tried to fine-tune the results obtained from the core system with humaninformed feature preference, and compared it with automatic feature selection as commonly practised in statistical WSD. Despite the insignificant improvement observed in this preliminary attempt, more systematic analysis remains to be done for a cognitively plausible account of the factors underlying the lexical sensitivity of WSD, which would inform and enhance the development of WSD systems in return.