Research On Question Answering System Based On Understanding Of Chinese Natural Language

The question answering system is an important research direction in understandingof natural languages in recent years, and also a practical application of the naturallanguage understanding. Based on research of the existing question answering system,this Paper takes the conceptual graphs as the main representation of knowledge to studythe structure, main algorithms and implementation efficiency of the question answeringsystem from understanding of the Chinese natural language.In research on the question answering system, the knowledge representation iscrucial. In order to enhance the abilities of semantic expression and knowledgereasoning, this Paper adopts the knowledge representation method of concept graphs tointegrate the representation of the simple and intuitive concept graphic knowledgefeaturing strong semantic expression ability into the traditional question answeringsystem to construct a question answering system based on Chinese natural languageunderstanding of the conceptual graphs.In the question answering system based on knowledge representation of theconceptual graphs, this Paper has mainly improved the performance of such threemodules as problem understanding, text search and answer extraction, and designed theNL-T-CG (Natural languages transform conceptual graphs) algorithm, the conceptualgraph semantic search algorithm and the conceptual graph clustering answer extractionalgorithm. First of all, in the NL-T-CG algorithm, the Hierarchical Hidden MarkovModel (HHMM) is adopted to achieve word-based treatment of the statements，then complete the lexical tagging and parsing, and design the conceptual graph generationalgorithm according to the word semantic relations and sentence relations in theanalysis results. Secondly, in the conceptual graph semantic search algorithm,preliminarily match the question with the relevant linguistic data by means ofconceptual graph projection matching according to the conceptual graph subject to theinput checking. Subsequently, check the structural similarity and context matchingfitness of the conceptual graph again, and then construct the conceptual graph semanticsearch algorithm according to the matched candidate answer set. Finally, in theconceptual graph clustering answer extraction algorithm, use the conceptual graphclustering method to recognize the answer type automatically. By sorting the matchingfitness to determine the answer extraction results, the algorithm integrates the intelligentalgorithm into the clustering algorithm to improve the process of conceptual graphclustering and design the answer extraction algorithm for conceptual graph clustering.On the basis of the above research, this Paper has achieved a question answeringsystem based on understanding of Chinese natural language, this system can offerqueries of Chinese statements to realize question understanding, document search andanswer extraction, and finally output a simple and intuitive natural language. Based ondebugging, experiment and test of the system, as well as comparative analysis of theresults after theoretical reasoning, this question answering system has achieved theaccuracy rate of51%, with its capacity increased by6%on average compared to thetraditional question answering. This system has already been applied to the projectresearch of Platform2.0cloud platform, and achieved satisfactory results.