Abstract: Semantic network is valid to semantic understanding and common sense reasoning,but coarse-grained semantic is difficult to represent complex relationship between objects．Though cognitive linguistics provides approaches to represent complex object-relational,it’s difficult to support natural language processing because of its abstract．This paper presented a method of natural language semantic representation based on cognitive linguistics．The method uses image schema to represent the relationship between object nodes in semantic network,and attribute spaces to represent the semantic of object which can be digitized(such as color)．On this basis,we transformed the building and modifying process on semantic network and attribute space into a sequence of semantic operations,so that we could reconstruct the semantics of sentences dynamically．Compared to the traditional method of semantic representation based on semantic network,the proposed method can express dynamic relationship and has stronger reasoning ability．The paper gave examples to prove the feasibility of this approach.

Key words: Cognitive linguistics,Semantic network,Attribute space,Image schema,Semantic operation,Reasoning