Research On Intelligent Question Answering System Based On Deep Learning And Self-attention Mechanism

With the continuous advancement of deep learning technology,the application in the field of intelligent natural language processing is becoming more and more common.The intelligent natural language question and answer processing system is an important achievement in the field of intelligent language processing,and it has gradually developed and expanded.However,the existing various types of intelligent natural language question answering systems still generally have excessive problems that rely on feature analysis engineering and statistical features based on word frequency,etc.,and cannot capture the contextual information of natural language text.Therefore,in response to the above problems,this article will conduct an in-depth study of the application of intelligent question answering systems by combining existing deep learning methods and self-attention mechanisms.The specific work is as follows:(1)Aiming at the problem of feature extraction,this paper designs a two-way long-term and short-term memory neural network model based on self-attention mechanism for feature extraction and is widely used in various intelligent question and answer tasks.The neural network model can effectively solve the problems of gradient feature disappearance and gradient feature explosion that traditional recurrent neural networks may encounter by adding three "gate" neural network structures;Features are extracted in both directions,so it can solve the problem of feature extraction context information and lack of context information.(2)The research results of this paper compare the application effects of several classic self-attention mechanisms on question answering tasks.In the learning model of deep learning oriented to intelligent deep question and answer,this paper introduces a method of combining deep learning with better performance and self-attention mechanism.By analyzing and calculating the feature importance and distribution vectors of different levels of words in the text vocabulary sequence,the important feature information of the words in the text is directly integrated into the intelligent question and answer model.The deep learning model in this paper has been extensively analyzed and verified on the basis of the public data sets Insurance QA and Web QA.The results show that the model based on the combination of deep learning and self-attention mechanism designed and studied in this paper is based on question and answer based learning tasks The application effect is superior to other deep learning-based algorithms.(3)This article uses the Tensorflow deep learning framework to implement the QA model in the article,and uses Java Spring MVC architecture and other technologies to design and implement a set of intelligent question and answer systems with low maintenance costs and easy deployment.