A Semantic Theory Based Upon Interactive Meaning

Abstract

When words are combined to form an expression, very often the expression means more than just the sum of the meaning of its constituents. The extra semantic information that the expression carries is semantic interaction of its constituents. Thus, the concatenation semantic features, words, or expressions which define the meaning of the individual constituents in an expression is not sufficient to represent the meaning of the expression. The semantic interaction of constituents has to be accounted for. In this paper, the problem of interactive meaning existing in noun-adjunct-noun constructions is discussed. It is shown that the current semantic theory proposed by the transformationalists is, as it stands now, not adequate in handling interactive meaning. A distributional theory of meaning is briefly presented in this paper. The meaning of a word is defined by the senses it realizes in different contexts and the conceptual relations of the senses to the senses realized by the other words. Concept-classes are formed on the basis of the distribution of word senses and the contextual restrictions of each sense are specified in terms of concept-classes. The theory is applied in this paper to resolve the problem of interactive meaning in noun-adjunct-noun constructions. It is suggested that semantic interactions between words are computable. A sample of interactive measnings in non-adjunct-noun constructions found in some restricted data and the rules necessary to determine them are presented together with a computational procedure to determine the interactive meanings of expressions in the construction.