Web Based Industrial Products Knowledge Graph Construction And Its Application

With the arrival of the internet plus age,both enterprises and individuals are more likely to rely on the internet recourse.At present,there are many resources about industrial products on the internet,including online encyclopedic,vertical websites,e-commerce sites,etc.How to fuse these different knowledge together and to draw a knowledge graph about various products as well as products and its attributes has become the focus problem for researchers.Due to the frequently update of the information about industrial products on the web and the content of knowledge graph also need to be constantly and dynamically updated.Therefore,in order to supplement the latest products information to the graph timely,web is used as the source of information to construct the products knowledge graph,and the unstructured text which contains the product information on the web is treated as the corpus resource in this paper.Then make research on entity recognition and relationship extraction among entities.For entity recognition,this paper proposed a hybrid model method to recognize product names,product attributes and attribute values.And according to the characteristics of product attributes,divide attributes into two categories,general attributes and special attributes.The rule-based method is adopted to recognize the general attributes and its values,and a two-stages method is adopted to recognize the product names,special attributes and its values.Experiments show that our hybrid model fuse the advantages of rule,post-processing of boundary,CRF and SVM,achieving satisfactory results in both effect and efficiency.For relationship extraction among entities,this paper takes sentence as a unit and extracts triples of product names,attributes and attribute values with pattern matching method.The extract results of this method shows a good reflection of the products information in the text through the constant enrichment and improvement the model we build.Finally,this paper designs and realizes the construction and application system of knowledge graph about the industrial products based on web.It integrates the information acquisition module and knowledge graph construction module,the former can obtain the information resources on the web at any time,while the later can extract the structured information in it and using the Neo4 j graph database to show them.This system also realize the application of product information visualization and related product recommendation.