Polysemy

  The fundamental problem posed by polysemy for computation is that particular words can take on an almost indefinite number of subtle meaning variations. These variations can stem from regular sense extension processes, be induced by contextual factors, or result from metaphorical or metonymic extensions. The lexicon is not fixed; word use is flexible and generative. This runs counter to the greatest part of NLP systems, which require every entry in the lexicon to be specified in advance and all potentially relevant sense distinctions to be permanently encoded. We saw in the previous section that for many applications a fixed view of the lexicon is adequate, but as more sophisticated processing of natural language becomes necessary the effectiveness of this approach to lexicon design quickly diminishes. It simply does not reflect the creativity which pervades word use. Let us consider some of the reasons why.

 
• Homonymy vs. Polysemy
  • Lyons
  • Cruse
  • Kilgarriff
• Perspectives on the representation of polysemy
  • Multiple Lexical Entries
  • Regular Polysemy
  • Underspecified Representations