Modern Chinese preposition is an important part of speech in Chinese grammar. Though preposition has no external meaning, it is the necessary semantic symbol in Chinese, which is adopted for word sense disambiguation, semantic analysis and so on. Therefore, preposition has a crucial role in the process of Chinese semantic understanding. The main purpose of research on prepositional understanding based on semantics is for semantic understanding of prepositional related topics. Based on the concept knowledge tree model, this thesis makes a comprehensive study of the preposition, includes prepositional knowledge database, prepositional related semantic model, prepositional concept related knowledge tree and prepositional automatic semantic analysis. The main contributions and novelties are summarized as follows. [1]Concept knowledge tree model is knowledge representation model based on concept. This thesis further studies on knowledge database of this model. [2]Based on achievements of Chinese linguistics and prepositional corpus, we built the first special Modern Chinese Prepositional Knowledge Database for Chinese semantic understanding, which consists of prepositional concept database and prepositional corpus. [3]This thesis summaries semantic roles of preposition signs based on the previous research, which are classified into 4 levels, 6 classes and 17 subclasses. And prepositional concepts are generated according to the defined semantic roles. We also sum up both syntactic and semantic characteristics of preposition analysis of corpus. [4]This thesis presents semantic model of preposition, prepositional phrase and its modifier. And preposition related knowledge trees are constructed. [5]Semantic understanding of preposition consists of automatic labeling the semantic role of prepositional sign and identifying prepositional phrase and its modifier. To solve these two problems, inductive logic programming and conditional random fields are adopted respectively. Experimental results show a promising performance of our methods.
修改评论