Knowledge-Based Methods for WSD
Rada F. Mihalcea · Kluwer Academic Publishers eBooks · 2006
This chapter provides an overview of research to date in knowledge-based word sense disambiguation. It outlines the main knowledge-intensive methods devised so far for automatic sense tagging: 1) methods using contextual overlap with respect to dictionary definitions, 2) methods based on similarity measures computed on semantic networks, 3) selectional preferences as a means of constraining the possible meanings of words in a given context, and 4) heuristic-based methods that rely on properties of human language including the most frequent sense, one sense per discourse, and one sense per collocation.