Crossing Parallel Corpora and Multilingual Lexical Databases for WSD

Word Sense Disambiguation (WSD) is the task of selecting the correct sense of a word in a context from a sense repository. Typically, WSD is approached as a supervised classification task to get state-of-the-art performance (e.g. [1]), and thus a large amount of sense-tagged examples for each sense of the word is needed, according to the word-expert approach. This requirement makes the supervised approach unfeasible for “all-words” tasks, consisting on disambiguating all the words in texts. This problem has been called the Knowledge Acquisition Bottleneck and many solutions have been proposed for it (see for example [2]) .