Research On The Construction Of Carbon Trading Knowledge Graph Based On Web Data

As an emerging technology,knowledge graph with powerful semantic processing and open organization capabilities is one of the current research hotspots.Knowledge graph can be divided into general and vertical domain knowledge graph according to different knowledge coverage.Currently,the research on knowledge graph is mainly concentrated on general domain,but the research on vertical domain is less.In recent years,Chinese carbon trading market has developed rapidly.It is urgent to propose effective integration methods to solve large number of heterogeneous and multi-source data.Traditional information integration methods cannot manage data from the perspective of knowledge.To address this problem,a method to construct a knowledge graph in the field of carbon trading is proposed in this thesis,which can integrate knowledge in the field of carbon trading from web data.This thesis focuses on the knowledge acquisition of knowledge graph and proposes a technical framework for constructing knowledge graph in the field of carbon trading.The main research contents and achievements are as follows:First,considering the characteristics of multi-source heterogeneous and distributed autonomous data in the carbon trading field,a data acquirer is constructed to automatically collect relevant data in the field of carbon trading from network resources.Second,different knowledge extraction methods are proposed for different structured data.For semi-structured data in the encyclopedia website,knowledge is acquired from the web data wrapper.For unstructured text data in vertical sites in the field of carbon trading,its knowledge extraction is divided into two parts: entity identification and relationship extraction.In the entity recognition phase,the BiLSTM-CRF network model is trained to identify entities in sentences.The average accuracy of it can reach over 90%.In the relation extraction phase,the method of pattern matching based on dependency syntactic analysis is used to obtain the relationships between entities,and a pattern generation method based on annotation data is proposed.The experimental results show that this method can effectively extract the relationships between entities.For the documents of industry standards and technical specification from the national technical standards websites,the industry knowledge can be extracted by building rules.Finally,the acquired triple knowledge is integrated and transformed into the form of linked data,and then a knowledge query module can be constructed based it.At the same time,knowledge visualization is realized through the graph database Neo4 j.The experimental results show that the knowledge graph construction method proposed in this thesis can effectively obtain triple knowledge from different structural data,and reduce human participation to a great extent;knowledge query and visualization presentation based on knowledge graph can provide assistance for knowledge services in the field of carbon trading.