形容詞のスケール性に関する多様性とレキシコン

Abstract

The aim of this paper is to reveal what some adjectives, e.g. Japanese adjectives aka-i (red) and naga-i (long), contain as lexical information in terms of scalability. The research has two steps: first, some parameters are advocated for describing each use of adjectives; second, these parameters are set in each use of a set of samples randomly extracted from the corpus. With regard to the first step, each parameter is constructed by the combination or arrangement of "values." They are minimum units which can vary at the level of perception or conception, independent of the linguistic expressions. As a result, the parameters pertain to: (i) the number of values, (ii) the way of extension of values, (iii) the number of linguistic expressions in a group of values, (iv) the number of "participant" in a group of values, and (v) the time flowing along values. The settings of these parameters are, of course, changeable according to the type of discourse, even in the same adjective, though each adjective has the tendency on selecting their parameters. Considering the second step, these parameters are applied to some specific adjectives of Japanese. Within a fixed number of samples including these adjectives, the parameters are set in each sample in reference not only to the sentence but also to the discourse. Eventually, it is found that each adjective shows the cluster of selection for parameters, which can be regarded as lexical information of the adjective. Since this way of analysis has some applicability to any adjective in any language, this makes it possible to compare adjectives in a language, or across languages.