문법관계와 한정성에 기반한 조사 분류－ 통계적 분석을 이용한 접근 －

Abstract

This paper aims to propose that each Korean marker can represent a proper grammatical relation, but not always. It is shown that quantitative patterns of Korean markers based on grammatical relations and definiteness, as found in the 21st Century Sejong Treebank, are estimated by a multivariate statistical method, Correspondence Analysis. An interesting observation is that (i) each grammatical relation tends to be concentrated by markers representing a proper meaning like hearer’s attitudes, exclusive selectivity, and so on; (ii) it is different to definiteness effects on grammatical relations reflected by markers. For examples, markers concentrating on subjects are sensitive to definiteness, whereas markers on objects have not sensitive to definiteness.