A Study On Automatic Generation Of Templete Library In Intelligent Question And Answering System In Automobiles Domain

With the high-speed development of Internet,the application value of the question answering system has attracted more and more attention by the industry,some enterprises have tried to apply answering questions to online QA,online customer service and so on.It is helpful to the integration of information,to reduce labor costs and to enhance customers experience.Question answering system is a collaboration of Nuartal Lnauggae Processnig,Inofmration Rertieval,Infomation Extraction and Knowledge Representation,which has been one of the most active research direction in the field of natural language processing.In this thesis,we study the automobile domain question answering prototype system which uses a template library,which associates with the problem expression,the answer way of query expressions,and the knowledge base of query.To a certain extent it transfers the complex logic of natural language processing to the process of the construction of the template library,which effectively improves the performance of question answering system based on template library.But the construction of the template library is a relatively laborious work,the studying of its automatic generation method can effectively reduce the construction complexity of the specific domain question answering system.To study the template library automatic generation method,the author focus on the characteristics of the question in the field of automobile,the construction strategy of knowledge base,the generation strategy of the query expression,the preservation,update,and iterative strategy of knowledge and so on.Study analyzes the characteristics of automobile brand models in question and realizes the car brand recognition algorithm of model;Study analyzes the difference between the property graph model and traditional relational model and their respective characteristics,and combines with the development of graph database and the advantages of the property graph model,and constructs the knowledge base and the semantic web,and studies the acquisition method of the semantic network related elements on the corpus of question;For synonymous expressions or wrongly written characters often appears in question,some methods based on word or Chinese Pinyin like the vector similarity,edit distance and Jaccard algorithm for fuzzy matching have been studied;For the classification of question in the field of automobile,this thesis uses machine learning technology,compared the classification effect based on bayesian classifier with decision tree classifier.Finally,this thesis introduces the structure of template library automatic generation system and related development tools,and proves the validity of the template library automatic generation method by some examples.And through the experiment of core function module,it demonstrates the performance and quality of the template library automatic generation method.