사용역에 따른 비윤리적 표현의 분포와 사용 양상 연구 : ‘방송 대화’와 ‘메신저 대화’를 중심으로

초록

This paper analyses various types of unethical expressions and distribution features in both large-scale broadcasting corpus and messenger corpus. The use of these unethical expressions appear to vary considerably depending on the register. As such, annotations for de-identification should be based on a register-specific approach rather than a general one. The results of the study can be summarized as follows. First, unethical expressions are categorized into four types: 'swearing expressions, hate speech, aggressive expressions, and sexual expressions.' Second, the quantitative analysis shows that the amount of unethical expressions in messenger is much higher than in broadcasting. Third, the proportion of [+person] expressions is very high in broadcast conversations, while swearing expressions account for more than 90% of the unethical expressions in the messenger corpus. Our study suggests that register variation, contextual information and language categories beyond word unit need to be reconsidered to detect unethical expressions.

키워드