Research On Question Classification And Answer Quality Prediction Method In Automotive Fault Question Answering Community

In the past,car owners always use search engines to make the preliminary judgment of automotive faults,but search engines are often difficult to accurately locate user information needs,leading to poor search results.Because of this,some professional automobile portal websites have successively launched question answering community,in which car owners can ask other users questions to obtain knowledge related to automotive faults,and independently complete the preliminary judgment of automotive faults,which not only meets the specific information needs of users,but also enhances the interaction between users.With the continuous accumulation of Q&A knowledge,how to effectively integrate information resources,use intelligent technology,quickly,efficiently and accurately provide users with high-quality Q&A information has become an urgent matter of the moment.In view of the above situation,this paper studies the design architecture of the intelligent management system of the automotive fault question answering library,and studies the corresponding algorithm for the two key tasks of automotive fault problem classification and answer quality prediction.The specific work contents are as follows:(1)Collect,process and analyze the question and answer data in the field of automobile fault,according to the characteristics of the question and answer data and the required functions of the system,design the intelligent management system framework of the question and answer library for the question and answer community,and integrate the question and answer resources of automotive fault effectively.(2)Aiming at the sparse text features and incomplete semantic information of automotive fault problems,a new classification method based on semantic co-occurrence of question-answer is proposed.Byusing the vector space model,words with similar semantics in question text and answer text are taken as co-occurrence words,and the attention mechanism is used to focus the features of co-occurrence words in question text,which effectively improves the accuracy of question classification.(3)Because of the shortcomings of the present method of answer quality prediction in measuring the semantic information of answer text,a method of answer quality prediction based on question-answer joint learning is proposed.Build Q&A respectively using attention mechanism of text representation,obtaining the dependent relationship between the two,reuse deep learning model to extract the Q&A on the semantic features,computing the semantic matching degree,and joint deep answer semantic representation and other valid extension features,said of the input as a full connection layer,effectively improve the quality of the answers prediction effect.This paper collates and marks the collected automotive fault Q&A data,constructs the automotive fault Q&A dataset consisting of 56,815 records,and improves and verifies the above two mission-critical algorithms.The experimental results show that the improved algorithm has good performance,which effectively verifies the feasibility of the design architecture of intelligent management system of automotive fault Q&A library for question answering community.