Earticle

Syntactic parsing is one of the most important elements in the machine translation systems of which base language is Korean, and it is one of the most difficult tasks in parsing Korean sentences to analyze the dependency among clauses. Though machine learning methods achieved high performance in many natural language tasks, they are difficult to be applied to analyzing clausal dependency since this task handles parse trees directly. To tackle this problem, this paper proposes a novel method which uses parse trees in analyzing clausal dependency. In this method, the structural information and the lexical information are separated, and a proper base kernel for each information is proposed respectively. Finally, a composite kernel combining both base kernels is used to analyze unseen cases. When the proposed method is applied to support vector machines, it shows 82.12% of accuracy for STEP 2000 corpus, which implies that the proposed method is plausible.