Research On Construction Of Automatic Question Answering System Based On Neural Network Semantic Matching

With the rapid development of Internet technology and the widespread application of big data technology,how to accurately obtain the true intentions of users in a large amount of network information has become increasingly important.How the automatic question answering system can understand the user's question and provide the information more intelligently and accurately for the user is a hot spot in the current academic and industrial research.In the current common method of answer selection,the quality of the question-and-answer match depends on the construction of sentence feature vectors.It is directly used to model sentences and it is easy to miss semantic information between words,which has a great impact on the calculation of matching degree.This article mainly addresses the shortcomings of the current mainstream answer selection methods based on improved comparative alignment semantic information answer selection techniques.The answer selection technique based on the improved comparison of aligned semantic information is divided into two steps: "coarse matching" and "fine matching".First,"rough match" the question and the answer.The keywords of the question are extracted mainly through the improved keyword extraction technology based on the three influences.In the convolutional neural network text classification model,the extracted keywords and key phrases are used instead of the original text for classification,and a candidate answer set of the category to which the question belongs is obtained.In the candidate answer selection,a comparative alignment semantic matching model is selected,and the framework's preprocessing layer,attention layer,and comparative alignment layer are improved,and the improved model is used to "fine match" the question and the candidate answer set to obtain a match The highest-scoring answer is the correct answer.The paper applies the proposed answer selection technique based on improved alignment and semantic information to an automatic question answering system based on the insurance industry for testing and verification.The system can extract the answer that is more in line with the customer's question intention from the existing database and successfully return To customers.Compared with the existing answer selection technology,the accuracy of the semantic matching of the answer selection technology of the system is improved.