Research And Implementation Of Government Procurement Question Answering System Based On Knowledge Graph

With the development of Internet information technology,information acquisition methods based on traditional search engines have been difficult to meet people’s needs for knowledge acquisition,and the intelligent question answering system arises at the historic moment.Different from traditional search engines,the intelligent question answering system uses natural language processing technology to obtain answers and return them directly to users,improving search efficiency.At the same time,knowledge graphs have developed rapidly in recent years.It provides a technical means for the upgrade of intelligent question answering system,which makes the working mode of question answering system closer to the way of human thinking.This paper starts the research work on the background of a company that has been working in the field of government procurement for many years and has realized the demand for dialogue and question-answering services of an electronic system platform in the whole process,time and field of procurement.The work of this paper mainly includes the introduction of government procurement knowledge graph technology in the question and answer service,and the combination of natural language processing technology and deep learning method to realize the government procurement intelligent question answering service.The main research contents of this paper includes the following four aspects:(1)Summary of question and answer service.Based on the actual needs of selected enterprises,this paper analyzes the problems of government procurement q&a system based on knowledge graph,and discusses the relevant technologies and algorithms.(2)Build a knowledge graph for the restricted field of government procurement.This paper proposes a knowledge graph construction method for government procurement.First,the paper draws on the seven-step method proposed by Stanford to define the knowledge model at the top of the construction process.After that,for the knowledge extraction task,the paper proposes a pipeline extraction model based on Bi GRU:(1)Use the Bi GRU-CRF model based on word mixture vector to extract entities;(2)Use the relationship classification model Bi GRU-Attention to obtain the relationship between entities;(3)The knowledge graph triples that constitutes the smallest unit.Secondly,for the knowledge fusion task,the paper uses the Word2 vec model to calculate the spatial distance between words,and completes the measurement of semantic similarity between words.Then,by setting the similarity threshold,the segmentation of the relationship between entities is completed,so as to fulfill the need of entity alignment.Finally,this paper introduces expert experience and knowledge to improve the quality of the knowledge graph,and stores the updated content in the neo4 j graph database.(3)Complete user question analysis and query Answer retrieval.This paper studies and applies the classification and answer retrieval technology of government procurement questions based on knowledge graph:(1)Use entity extraction model to identify the question entity;(2)Link the question entity with the corresponding entity in the knowledge graph;(3)Text CNN-Attention-based is the question classification model to recognize the intent of the question,parses the natural language question,and converts the natural language question into a Cypher statement for querying in Neo4j;(4)With the help of Cypher query technology,search and push the answer in the graph database.(4)Design and implement an automatic question answering system for government procurement based on knowledge graph.Taking html+query as the front-end framework,python+Sanic as the back-end framework,using python language,based on the research results of the thesis,developed the government procurement question and answer system software,and tested the software according to the software engineering method,and the test results proved that the project design was realized target.