Spatial Structure Analysis and Identification of Candidate Hyponymy Relations

Automatic acquisition of semantic relations is an important problem in knowledge acquisition. We present a method of spatial structure analysis of candidate Chinese hyponymy based on concept space and use them to identify candidate hyponymy. Firstly a group of candidate hyponymy relations is imported into concept space. Secondly we analyze the basic spatial structure of concept space including concept nodes and relation edges. More identified features based on the spatial structure analysis are given and used to identify hyponymy relations. Experimental results show that the spatial structure is very useful to the identification of hyponymy.