Design And Implementation Of A Knowledge Graph Platform For The Textile Industry

Knowledge graphs(KGs)are a class of semantic networks that abstract and model real-life concepts such as entities,attributes and relationships to fuse different information into a unified graph structure.Knowledge graphs can be divided into general knowledge graphs and domain knowledge graphs.The former is oriented to open domains and general scenarios,such as various kinds of encyclopedic knowledge graphs;the latter is oriented to a specific industry or proprietary domain.Industry knowledge mapping belongs to one kind of domain knowledge graphs,which aims to carry out deeper information mining from the perspective of industrial chain and industrial structure by modeling entities and relationships related to a certain industry,so as to optimize industrial chain and structure and improve industrial efficiency and production level.This paper collects data from public industrial and commercial information on the Internet and textile industry encyclopedia,model from the perspective of industry chain,build textile industry knowledge map,mine the implicit information of textile industry,and display relevant information in a user-friendly way to provide reference for industry development.This paper focuses on the construction method of industry KGs and the representation learning method for industrial chain,with the goal of constructing a high-quality industrial knowledge graph and graph node representation at a low cost and forming a method that can be extended to other industries.In addition,this paper designs and implements a knowledge graph platform for textile industry through relevant technologies and tools,and finally completes the deployment.The main work of this paper is as follows:(1)Designing a framework for building knowledge graphs about industries.By incorporating the encyclopedic domain and the industrial domain into the industry knowledge graph construction,the knowledge quality and completeness of the industry knowledge graph are improved at a small cost,and the sufficient data foundation is also provided for the subsequent graph representation learning.(2)Construction of textile industry knowledge graph.This paper uses Selenium-based automated testing tools to collect data from publicly available Chinese encyclopedia websites on the Internet and enterprise industrial and commercial information websites,and design ontology models from the perspective of textile industry chain and persistent storage of knowledge graphs based on Neo4 j graph database.(3)Knowledge graph node representation learning based on graph comparison.In this paper,the node features are initialized into text classification task by using pre-trained language model through industry chain classification,and then the topological structure information of nodes for graph structure is further incorporated through GNN,and in order to alleviate the anisotropy problem of pre-trained language model BERT,graph comparison learning is introduced to obtain the information incorporating semantic information,topological information,and inter-node comparison information to learn higher quality graph node representations.(4)Design,implementation and deployment of a knowledge graph platform for the textile industry.Based on React.js and Fast Api framework,this paper designed and implemented the textile industry knowledge mapping platform to provide textile industry knowledge mapping query and management functions.Use Docker to divide and configure containers for each functional part of the platform,and realize the deployment for the platform.