남성과 여성 발화의 어휘적 차이: 코퍼스 기반의 남녀 판별 연구

Abstract

The present paper seeks to find lexical differences of men and women in spoken utterances on the basis of a spoken Korean corpus composed of 350 thousand words. First, a new interpretation, based on formality, is given to the result of a previous experiment(Han and Kang 2009). Next, new experiments of automatically classifying a set of utterances as men's or women's are conducted. The success rates are as high as 80%. In the meantime, by way of chi-square statistic some characteristic words of either men or women can be listed. The list itself is an interesting linguistic observation, which are interpreted in various ways including semantic prosody. When we use only small number of words showing most prominent gender characteristics, the success rates of the discrimination experiments are not lowered compared with the case when we take all the words into account. The results show that considering only small number of gender-oriented words is enough (even better) for identifying male or female utterances.