Semantic Labeling of Chinese Verb-Complement Structure Based on Feature Structure

Abstract

Semantic relations are difficult to analyze automatically, especially the semantic relations of Chinese verb-complement structure. In this paper we propose a novel model based on feature structure and apply it to the representation of semantic relations among subjects, verbs, objects, and complements. We focus four different kinds of Chinese verb-complement structures and their semantic relations. We compared the approach we propose with traditional dependency grammars. Feature structure, being recursive undirected graph, facilitates a richer Chinese semantic information extraction when compared to dependency grammar. The results of our analysis show that using feature structures are more suitable for extracting complex semantic relations.