Abstract

this paper [3], Justeson and Katz argue for a "linguistically principled" approach to disambiguation in which only syntactically or semantically relevant clues are used. Statistical methods are used to organise and analyse data in "training", but the disambiguation stage does not use statistical techniques. The actual system they describe disambiguates adjectives using only the nouns that are modified by the adjectives. Initially, the system can be completely automated, trained using statistical techniques to pick out nouns which indicate the sense of the adjectives concerned. In order to provide greater coverage the system is later implemented using hand-coded rules rather than the statistical information. Justeson and Katz performed all testing on "five of the most frequent ambiguous adjectives in English": "hard", "light", "old", "right", and "short". 2.3.1 Required data and systems

Citations