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 “9 Comparison with previous work” · whole deck

9 Comparison with previous work

  1. Why is bilingual (parallel) data an imperfect alternative to hand tagging for WSD?
    Aligned bilingual corpora are of limited availability, and translation differences often fail to model monolingual sense differences.
  2. How does Yarowsky’s approach improve on dictionary-based WSD such as Lesk’s?
    Earlier methods measured overlap with dictionary definitions; Yarowsky combines the limited seed information in definitions with the nearly unlimited co-occurrence information in corpora.
  3. How does Yarowsky’s algorithm differ from Schütze’s sense clustering?
    Schütze clustered contexts without supervision and aligned clusters to senses afterwards by hand, using bag-of-words documents. Yarowsky ties classes to the desired senses from the start with seeds, and models collocational distance, word order and predicate–argument relations.
  4. How did Hearst’s earlier bootstrapping differ from Yarowsky’s?
    Hearst started from hand-labelled sentences and added confidently tagged examples (described in two sentences, not developed further). Yarowsky needs no hand-labelled data and jointly uses collocation and discourse constraints.

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/