Word Sense Disambiguation

David Yarowsky

Word sense disambiguation (WSD) is essentially a classification problem. Given a word such as sentence and an inventory of possible semantic tags for that word, which tag is appropriate for each individual instance of that word in context? In many implementations these labels are major sense numbers from an on-line dictionary, but they may also correspond to topic or subject codes, nodes in a semantic hierarchy, a set of possible foreign language translations, or even assignment to an automatically induced sense partition. The nature of this given sense inventory substantially determines the nature and complexity of the sense disambiguation task. After surveying applications of WSD, this chapter discusses the extraction of contextual evidence sources and their use in supervised learning algorithms for word-sense classification. Lightly supervised and unsupervised methods for sense classification and discovery are also described for situations when costly hand-tagged training data is unavailable or is not available in sufficient quantities for supervised learning.

Bibtex Citation

  @incollection{yarowsky-handbook10,
    author = {David Yarowsky},
    title = {Word Sense Disambiguation},
    booktitle = {Handbook of Natural Language Processing, Second Edition},
    editor = {Nitin Indurkhya and Fred J. Damerau},
    publisher = {CRC Press, Taylor and Francis Group},
    address = {Boca Raton, FL},
    year = {2010},
    note = {ISBN 978-1420085921}
  }