Automatic Construction Of Knowledge Database And Question Searching For Specific Domain Question Answer System And System Implementation

With the rapid growth of network information,people expect to obtain information in a more concise way.The traditional search engine can only return a series of web pages,people look forward to obtain information through a more concise way,the question and answer system came into being.Question answering system can be roughly divided into two categories: open domain question answering system,restricted domain question answering system.There is a great difference between the actual working process of these two kinds.Specific domain question answering system is generally divided into three parts: the construction of knowledge base,the analysis of user question,question retrieval.This paper focuses on the construction of knowledge base and question retrieval of specific domain question answering system.The building knowledge base involves three aspects: the choice of data sources,the way of knowledge organization,and the degree of automation.Through the research,we found that most of the research work in the selection of the data sources,often only consider the Wikipedia,the question and answer community,the field sites,etc..Few consider the use of multiple data sources to build knowledge base.In addition,many research work is based on ontology as knowledge organization,and then according to specific areas,specific methods are used to build a specific knowledge base.Because of the particularity of ontology,the method of constructing the knowledge base is difficult to transplant,once the field or demand changes,you need to start building a knowledge base,past work can hardly reuse.Therefore,in order to improve the automation level of building knowledge base of the specific domain question answering system,this paper proposes a framework for the automatic construction of knowledge base for multidata sources.The framework use the field sites,wikipedia and the q & a community of data sources as a source of knowledge,and use question and answer pair as the main organization.The data in this thesis based on domain specific domain terminology to collect encyclopedia and answers community two knowledge sources,resulting in a domain term precision will be directly related to the accuracy of the knowledge base,so the domain term extraction work done for further study,and the term extraction algorithm based on word2 vec is improved.In the retrieval,because of restricted domain question retrieval framework only for the local domain knowledge base of traditional retrieval,when unable to find a suitable candidate question,will not deal with questions on the part of the user's questions.So therewill often happen "no reply" phenomenon.In order to alleviate this problem,this paper proposes to combine domain knowledge base and a number of online q&a community question retrieval framework.In addition,in the restricted domain question retrieval framework process improved,adding automatic expansion of the function of knowledge base.Finally,based on the previous research work,this paper proposes a set of strategies based on the wechat platform to build a limited domain question and answering system.The strategy is universal,users only need to follow the instructions of the configuration file,then quickly set up a specific domain question and answer system.In addition,all the work in this paper has been on the github open source,hoping to provide help for other researchers.