Research On The Construction Technology Of Knowledge-Based Question Answering System

With the improvement of people's living standards,many people take travel as part of their spiritual life,but in the era of big data,the Internet is full of complex travelling guideline information,and traditional search engines will return a large number of web pages to tourists.Tourists also spend a lot of time extracting the information they need,and the traditional search model has been increasingly unable to meet user needs.How to return accurate information to users is an urgent problem,the knowledge-based question answering system is one of the methods to solve this problem.The knowledge-based question answering system provides users with an efficient and accurate way to get information,it can return short and accurate answers to users.Therefore,constructing a tourism knowledge-based question answering system can effectively solve the problem that tourists need to spend a lot of time in collecting information.Therefore,this paper mainly studies three important technologies used in the construction of the tourism knowledge-based question answering system: techniques for building a knowledge base,extracting the subject entity of the question and the attribute of the question.The specific content is as follows:1.Building a travel knowledge base to integrate travel information from various websites.The general domain knowledge base on the Internet cannot meet the daily research needs of specific domains,so this article takes Guangxi as an example to construct a knowledge base of tourist attractions.2.Entity recognition based on entity storage network.To solve the problem that the sentence features extracted from the datasets by many neural network-based models are semantically incoherent.This article proposes an entity storage network model to store the extracted local language features and then combines the attention mechanism to expand the scope of local linguistic features.Experimental results show that the entity recognition method proposed in this article has achieved good results on two different fields of datasets3.Question attribute extraction based on information interaction matrix.Aiming at the problem that mainstream methods only consider high-level semantic information and ignore word-level text information,this paper presents a method of attribute extraction by constructing interaction matrix of question and attributeinformation.Questions contain part of information similar to attributes at the text level.This article constructs an information interaction matrix of questions and attributes,and then extracts the interactive information by convolution.Finally,combining the interactive information with high-level semantic information to choose the most appropriate attribute from the knowledge base.Experimental results show that the attribute extraction method proposed in this paper has achieved good results.