Abstract:
Most previous studies on “good/bad” are qualitative while corpus-based quantitative studies are rare. The present study investigates the asymmetry between “good/bad” in terms of semantic and usage feature using a corpus-based behavioral profile analysis. A hierarchical agglomerative cluster analysis and correspond analysis were performed. The results are as the following: 1) the asymmetry between good/bad is reflected in the semantic extension: the literal meaning of “good” extends to temper, character, degree or range, idea and behavior, while that of “bad” extends to temper, feeing, state, language, behavior and character. 2) the asymmetry between extended meanings of “good” and “bad” is influenced by two factors: head nouns modified by “good/bad”, subjects of clause, while the rest two factors, morphological and polarity features, are only the one of reasons causing differences between literal and extending meanings. In addition, the tense has nothing to do with asymmetry. Combining cluster analysis with correspond analysis, this study provides a new path for the study of antonym asymmetry and further demonstrates the applicability of this empirical method, meanwhile gives support for relevant research of teaching application.