Word Sense Disambiguation

tion from English to German, needs to translate "bank" as "Ufer" for sense 1a and as "Bank" for sense 1b. An information retrieval system answering a query about "financial banks" should only return documents that use "bank" in sense 1b. Whenever a system's action depends on the meaning of the text being processed, disambiguation is beneficial or even necessary. Although we only cover statistical approaches here, work on word sense disambiguation has a long tradition in Artificial Intelligence and Computational Linguistics. For example, Kelly and Stone (1975) constructed a rule-based disambiguator, Hirst (1987) exploited selectional restrictions for disambiguation, and Cottrell (1989) built a connectionist model integrating syntactic, semantic and other constraints. However, none of these models addressed the question of training which is at the heart of the material we aim to present in this book. The nature of ambiguity and disambiguation changes quite a bit depending on w