초록

용어보기논문에서 용어와 풀이말을 자동 추출한 결과로, 시범 서비스 중입니다. AI-Helper

Abstract ▼ AI-Helper

This paper presents Korean dependency parsing algorithm using sets of head candidates. In this algorithm, the characteristics of Korean language has been used, which are agglutinative, head-final, and projective. Our approach differs from the general non-deterministic dependency parsing that searche...

This paper presents Korean dependency parsing algorithm using sets of head candidates. In this algorithm, the characteristics of Korean language has been used, which are agglutinative, head-final, and projective. Our approach differs from the general non-deterministic dependency parsing that searches whole dependents and heads globally. Instead of considering whole dependents and heads, we restrict to only heads which could be related in grammar for each dependent. Such an approach can reduce the parsing complexity. In addition, we have modified the structure of Sejong tree-tagged to dependency structures, which was originally composed of phrase structures. Such a tagged corpus has been used for training and test data. Our experiments show 87.52% arc accuracy and 34.43% sentence accuracy, respectively. Furthermore, the proposed algorithm shows 3 to 11 times faster than CYK parsing algorithm depending on the lengths of sentences.

주제어