Research On Intelligent Solution Technology Based On Neural Network And Knowledge Graph

With the continuous progress of electronic and computer technology,non-artificial solution technology has been widely used in many industries.Including e-commerce customer service,operator self-service navigation,and some consultation and answers in telephone advertisements,etc.,have been replaced by non-artificial question-and-answer systems.Compared with human customer service,the non-human Q&A system will undoubtedly significantly reduce the employment cost of enterprises.However,due to some characteristics of language itself,such as the multiplicity of semantics and the multiple ways of expressing the same thing,language is ambiguous.Therefore,in many question answering systems,the accuracy of information extraction and semantic understanding of questions is difficult to achieve satisfactory results,which greatly reduces the user experience.The document retrieval method(FAQ)based on preset question answering is the basic method of most question answering systems.This method is equivalent to assisting users to search for preset questions,but has poor flexibility and cannot handle personalized questions.In order to solve the pain points of the above answering technology,this paper uses the intelligent answering technology based on neural network and knowledge graph,and focuses on the exploratory research and implementation of the information extraction problem of natural language processing in the answering technology.The main content of this article is as follows:1.In order to solve the above problem of low answer accuracy and poor performance,this paper uses the answer technology based on neural network and knowledge graph to improve,proposes the Sparse SA-GCN model,and uses the sparse self-attention method to optimize GCN.The model reduces computing power consumption and improves the accuracy compared to some traditional benchmark models.2.This paper constructs a joint extraction model that integrates entity recognition and relation extraction,and solves the problem of the correlation between error accumulation and lost information relation in the pipeline information extraction model.3.According to the above methods,this paper implements a demonstration of intelligent answering based on neural network and knowledge graph,and uses the model to build a graph question answering demonstration of basic functions.The trained model can flexibly and intelligently answer the user’s differentiated questions in the field limited by the data set in the form of knowledge graph.Compared with the traditional answering method,it has two advantages of convenient knowledge base construction and high query accuracy,which has certain practical application value in real life.