Construction And Application Of Government Procurement Information Knowledge Graph For Small And Medium-sized Enterprise

With the rapid development of the Internet,the government procurement model has shifted from traditional manual procurement to electronic procurement.Although electronic procurement saves procurement costs,improves procurement efficiency and achieves information sharing.However,due to the huge quantity,variety and complex content of electronic procurement information,it also increases the difficulty of obtaining target information objectively.At the same time,due to the low degree of informatization of small and medium-sized enterprises,the way of obtaining government procurement information manually is not only inefficient,but also has poor information integration capabilities.Therefore,small and medium-sized enterprises urgently need an effective knowledge graph tool to obtain information related to their own interests from a large amount of government procurement information.Knowledge graph can quickly obtain effective information from massive Internet information,and efficiently construct the association between information in the form of graph structure.This thesis combines the knowledge graph with government procurement,and uses the knowledge graph to construct the government procurement information network to realize the integration of government procurement information.In order to apply knowledge graph to government procurement information,the following research work is done in this thesis:(1)The construction of government procurement information knowledge graph.According to the data characteristics of government procurement information,using domain ontology construction technology,natural language processing technology and deep learning technology,this thesis proposed a method for automatically constructing the knowledge graph of government procurement information.Firstly,the method uses the scrapy framework to crawl the procurement data of government procurement related websites.Then,use the scrapy framework to crawl data from government procurement related websites.Then,aiming at the characteristics that some basic entities have no obvious data characteristics,a basic entity extraction method based on BERT-Bi GRUATT-MRC model is proposed and tested on the procurement text dataset.The test results show that the F1 value of the model is 88.05%.In view of the characteristics that implicit entities cannot be directly obtained from the procurement text and the data range is relatively fixed,it is regarded as the text classification task,and the BERTText CNN model is used to extract implicit entities.In the content type and industry type classification tasks,the F1 value of the model are 93% and 91%.Finally,the extracted entities are stored in the Neo4 j graph database,and there are currently 110000 nodes and 150000 relationships in Neo4 j.(2)An automatic question answering method for purchasing information based on knowledge graph.Aiming at the problems of inaccurate word segmentation,unable to understand semantics and high labor cost in traditional template matching and semantic parsing knowledge question answering methods,an automatic question answering method for procurement information based on deep learning is designed.The method uses the BERT-WWM-Bi LSTM-CRF model to identify the subject words in the questions,and is tested on a dataset of 5146 questions.The test results show that the F1 value of the model is 90.09%.Then the Sentence-BERT model is used to identify the intention in the question,and test it on the 5126 relational similarity dataset.The test results show that the ACC value of the model is 90.23%.Finally,the method is evaluated manually,and the answer query accuracy is 80.25%,which verifies the effectiveness of the method.(3)Design and implementation of government procurement information knowledge graph management system.In order to facilitate users to manage and use the government procurement information knowledge graph,this thesis develops a set of government procurement information knowledge graph management system through system demand analysis,outline design and detailed design,and shows the main functions of the system to verify the practicability and feasibility of the system.