Research On The Fusion Of Multi-source Interpretations Based On Conceptual Graph

With the rapid development of Internet and the explosive growth of the information from the web, the CIR(Chinese Information Retrieval ) based on the boolean model cannot meet people's needs and the CCIR(Chinese Conceptual Information Retrieval) based on the Chinese Semantic Analysis is a new development point. The CCIR based on the CG (conceptual graph) includes the demand concept analysis,the CG annotation and match and the CG annotation can significantly affect the performance of retrieval systems. In order to improve the accuracy of CG annotation and check results the author intends to create an ideal conceptual knowledge base.The author chooses the machine-readable dictionary as a corpus source. There are some problems in the traditional DIE(Dictionary Information Extraction): first, the traditional DIE mostly chooses only one dictionary as the corpus source and the knowledge is not sufficient; secondly, the traditional DIE mostly focuses on depth of extraction rather than breadth. So, the paper will give a new method to create a ideal knowledge base.The work of the author is based on CG. The main contents of the work have three points:First, try to complete extraction and fusion of the internal attribute information from the Modern Chinese Norm Dictionary and the Modern Chinese Dictionary and get the Entity-Attribute Value pairs which comprehensive knowledge of the two dictionaries. It can be directly applied to CG annotation and verification to prevent further conceptual analysis of the dictionary and improve the accuracy of CG annotation and efficiency.Secondly, try to classify extraction patterns by clustering. By clustering extraction patterns are divided into different clusters and then according to different clusters the different attribute values are extracted from the dictionary. It greatly reduces the human consumption in recognition of extraction pattern.Thirdly, try to introduce the semantic similarity computation into clustering. By calculating the semantic similarity of the context the misclassification rate can be reduced and the extraction accuracy can be improved.Anyway, the paper tries to study the extraction and fusion of the internal attribute information from the Modern Chinese Norm Dictionary and the Modern Chinese Dictionary in order to supply some basic resources and ideas for CG annotation and CCIR.