learn.ifs.name

Unsupervised Word Sense Disambiguation Rivaling Supervised Methods

Yarowsky, 1995, Unsupervised Word Sense Disambiguation Rivaling Supervised Methods (Proceedings of the 33rd Annual Meeting of ACL)

Subscriber deck — free during early access

4 cards in “3 One sense per collocation” · whole deck

3 One sense per collocation

  1. How does the strength of one sense per collocation vary?
    It is strongest for immediately adjacent words and weakens with distance; stronger for predicate–argument relations than arbitrary words at the same distance; and much stronger for content words than function words (over 97% for adjacent content words).
  2. How does a decision list classify a word sense?
    Collocational features (lemmas, word forms, POS, word classes at various positions: adjacent, in a ±k window, trigrams, predicate–argument) are ranked by \(\left|\log \frac{P(\text{Sense}_A \mid \text{Collocation}_i)}{P(\text{Sense}_B \mid \text{Collocation}_i)}\right|\); a new example is classified by the single highest-ranked feature present in its context.
  3. Why use only the single most reliable piece of evidence instead of combining all?
    It avoids modelling the complex dependencies between highly non-independent evidence sources, which suits the many overlapping collocations available in WSD.
  4. Why is smoothing crucial for decision-list log-likelihood ratios?
    Most ratios involve a zero count for some sense; the smoothing depends on the collocation type, distance, content vs function word and expected noise.

Source

  • David Yarowsky. 1995. Unsupervised Word Sense Disambiguation Rivaling Supervised Methods. In Proceedings of the 33rd Annual Meeting of the Association for Computational Linguistics, pp. 189–196. https://aclanthology.org/P95-1026/