Research On Conceptual Graph Building And Matching For Chinese Semantic Based On E-A-V Structures

Semantic analysis is an important task in the research of natural language processing.Because there is no strict correspondence between the Chinese syntactic component and partsof speech, so the traditional lead-based semantic analysis method is not suitable for theChinese semantic analysis. Conceptual graph is a semantic knowledge representation whichhas been proven superior to the classical artificial intelligence methods. The concept semanticwhich using the conceptual graph will provide new ideas for the research of Chinese semanticinformation processing. Since Chinese words are morphemes words, they take the nouns ascenter part and often expressed by way of semantic through coupling between words toexpress the concept of direct coupling between the associations. So it is necessary to study theChinese semantic conceptual graph based on the traditional conceptual graph. In this paper,firstly, the traditional conceptual graph which is on the basis of the model proposed and theconcept semantic representation model. And then taking the strategy of weakening thesyntactic analysis and focusing on the syntactic analysis difficulties when faced theconceptual intensional analysis proposed for the Chinese intensional semantic conceptprocessing which based on the concept intensional analysis. In the end, it is to propose a newconceptual graph matching algorithm aimed to the problem of the inaccurate conceptual graphmatching. The main work of this paper is as follows:1. For the problem of the Chinese syntactic analysis, taking the strategy of weakeningsyntactic analysis and laying emphasis on the conceptual analysis, the paper proposed arepresentation model of intensional semantic conceptual graph what is suitable for Chinesesemantic analysis.2. On the basis of the representation model of intensional semantic conceptual graph, thepaper proposed the generation process of Chinese intensional semantic conceptual graph. Firstof all, analyze the basic concept to generate the basic conceptual graph. And then using thecenter concept of composite structure as the core and merging the basic conceptual graph, thecomposite conceptual graph could be generated. Finally simplify the composite conceptualgraph.3. In Chinese semantic computation, the conceptual graph matching had the inaccurateresult, so the paper proposed the algorithm of conceptual graph matching based on“Entity-Attribute-Value” structures. Based on the intensional logic model of Chinese conceptconnotation, using intensional semantic conceptual graph as knowledge representationmethod and combining with the computation method of E-A-V structures similarity, the algorithm computed the similarity of conceptual graphs. The experimental results show thatthe algorithm is more effective on conceptual graph matching.4. In the conceptual graph matching algorithm that based on the Entity-Attribute-Valuestructures, the conceptual structure similarity was generated by the average value of theEntity-Attribute-Value similarity. For this problem, there came out the strategy of conceptualsubgraph weight self-adjustment in the matching algorithm. When using this algorithm tocompute the conceptual graph similarity, it can give the homologous weight to the subgraphbased on the proportion of how much information the subgraph contains in the wholeconceptual graph. The experiment results show that this new algorithm achieve better results.