Information Extraction And Knowledge Graph Construction For Enterprises

With the establishment and rapid development of the China(Shanghai)Pilot Free Trade Zone(FTZ),the scale of enterprises in the region has grown rapidly,which has increased the difficulty of supervision and service of enterprises in the zone by the management departments of the Free Trade Zone.The data and information in its existing platform are not rich enough to comprehensively supervise and serve enterprises.Therefore,Internet big data and related technologies are urgently needed to support and strengthen the comprehensive supervision and service to enterprises.With the rapid development of the Internet,the scale of data related to enterprises on the Internet is constantly expanding.Therefore,extracting key information from massive enterprise big data helps the regulatory authorities to better grasp the dynamics of the enterprise.Based on the commissioned projects and actual needs of the Shanghai FTZ,this paper extracts key information such as entities(person name,location name,enterprise name,date)and relationships(enterprise and enterprise,person and enterprise)from the relevant big data of Internet companies.Further build a enterprise knowledge graph and apply it to the supervise and service of the FTZ.The main work of this paper is as follows:(1)Aiming at extracting key information of person name,location name,enterprise name and date entity,a joint named entity recognition method based on dictionary and deep learning is proposed.This method uses both dictionary-based forward maximum matching and deep learning based BiLSTM-CRF performs named entity recognition,and the final entity annotation result is obtained by comparing the two annotation results.In addition,we align the short name of the company name with the full name when comparing the entities.(2)For how to extract relationships from entities,including extracting the relationship between enterprise and enterprise,and the relationship between person and enterprise,the BiLSTM-BiGRUAttention model based on deep learning is proposed to extract predefined relationships,which can fully consider the context,capture key features,and extract relationships between entities.(3)Based on the above work,to better structure the entity and relationship information,use the Neo4 j graph database to build the enterprise knowledge graph.After the construction of the enterprise knowledge graph is completed,the relationship evolution algorithm is proposed by using the existing relationship rules.Finally,the enterprise knowledge map deployment is applied to the project to realize the application oriented to the FTZ.In order to verify the validity of the named entity recognition and relationship extraction methods proposed in this paper,the experiments have been verified and have achieved good results.Further use of entities and relationships to successfully build a enterprise knowledge graph and apply it to the comprehensive supervision and service of the FTZ management department.