영화 리뷰 코퍼스 MUSE-MOV에 나타나는 ‘명사(+조사)+<없다>’ 연쇄의 극성 양상 분류 연구
신동혁
(한국외국어대학교 박사과정 수료)
남지순
(한국외국어대학교 교수)
This study aims to analyze the polarity orientation of ‘Noun(+Particle)+’ sequences and to formalize the result as a Finite-State Transducer(FST) lexicon, named Local-Grammar Graph. ‘Eopta’ is a Korean negator, which usually shifts the basic polarity expressed by the collocated noun phrases. However, among the 1,123 sentences that contain ‘eopta’, extracted from the movie review corpus MUSE-MOV, about 76% occur with noun phrases without polarity. Seven sub-types of noun phrases co-occurring with ‘eopta’ have been discerned in this study. According to the semantic and lexical properties of these nouns, certain sequences of ‘Noun(+Particle)+’ have turned out to be positive or negative ones, which hardly may be predictable by rules. To automatically annotate these polar properties into each sequence ‘Noun(+Particle)+’ in the corpus, we constructed a FST lexicon that represents these 7 types of ‘Noun(+Particle)+’ sequences through Unitex platform and evaluated the performance of this lexicon via DecoTex platform. Compared to the results based on the dictionary of simple words, those based on our FST lexicon appear to be 30% more accurate. (Hankuk University of Foreign Studies)