Abstract
Abstract : This is a pilot study concerned with developing an inductive method for semantic classification of natural language forms. Consideration is restricted to forms expressing relations between entities or sentence elements. A paraphrasing technique called predication-typing is used to generate the data. This technique consists in paraphrasing sentences or portions of sentences, patterning the paraphrase according to a standard schema so as to incorporate one of a restricted set of expressions called 'predication types' as the connective element between the entities of the relation. The original natural language forms which have prompted selection of particular predication types are called cue forms. They may be then characterized in terms of the variety and frequency of selection of predication types whose use they have prompted in the course of processing a significant quantity of text. This characterization of a cue form in terms of predication types is called its 'semantic spectrum.' 'Semantic groupings' are derived from the semantic spectra which give promise of allowing for a systematic representation of semantic relations between natural language forms in terms of the meanings of the set of predication types as well as in terms of the two semantic dimensions of generality vs. specificity and vagueness vs. definiteness of meaning. (Author)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have