Exploiting Rules for Word Sense Disambiguation in Machine Translation

This paper describes the automatic generation and the evaluation of sets of rules for word sense disambiguation (WSD) in machine translation. The ultimate aim is to identify high-quality rules that can be used as knowledge sources in a relational WSD model. The evaluation was carried out both automatically, by means of four objective measures (error, coverage, support and novelty), and manually, by means of a subjective analysis of the level of interest of the best rules as pointed out by the objective measures. As a result, we selected 63 rules addressing seven highly ambiguous verbs. The evaluation also evidenced which kinds of knowledge were effectively used by the WSD rules, which are not always the same as those revealed by traditional evaluations of complete WSD models. Although we experimented with English-Portuguese, the rule generation and evaluation procedures could be applied to any language pair, provided that there is a disambiguation sample corpus for that language pair.