The Construction And Application Of Knowledge Graph Incorporating Causal Events

The knowledge graph integrates the knowledge resources of the Internet in Shanghai in a networked way to form a huge semantic network,providing a better ability to organize and utilize information,and has developed into the infrastructure of the big data era.With the development of information extraction technology and the needs of practical applications,people began to mine event knowledge from unstructured texts.Event knowledge is an important decision-making basis.Everyday,there are various events happening.These events usually do not exist in isolation,and there is often some semantic causal logic.The causal logic between events is a very valuable knowledge.The knowledge of causal events is supplemented by the knowledge graph,which can further enrich and improve the knowledge graph,so that the knowledge graph can play more value in practical application.Taking the financial field as an example,this paper firstly constructs a preliminary financial knowledge graph based on semi-structured data,and then extracts causal event knowledge from unstructured financial news and integrates it into the financial knowledge graph,further enriching and perfecting it.The financial knowledge graph is used in application scenarios such as analysis and decision making based on knowledge graphs.This article has mainly completed the following work:The framework of the financial knowledge graph is proposed.Based on the detailed research and analysis of the concepts and knowledge in the financial field,the important concepts and classes in the financial field are abstracted out,and the attributes and values of the entities are defined in detail,completing the financial The construction of ontology library;at the same time,the problem of redundant information appears in the process of mapping relational database to RDF graph by D2 R tool.This paper proposes a data table design principle,which separates the entity and entity relationship into data table storage.This problem is solved very well;then the knowledge extraction is completed using D2 R tool;finally,the knowledge graph is stored in the Neo4 j graph database,and the preliminary financial knowledge graph is constructed.The sequence annotation method is used to transform the causal event extraction task into a sequence annotation task,which realizes the extraction of causal events from unstructured financial news.The experiment compared three different sequence labeling models and found that the Bi-LSTM+CRF model has the best performance,and the F1 on the test set reached 79%.Then,using the model,a total of 5278 causal event pairs were extracted from the news.The "cause events" and "results events" in the causal event pair were respectively treated as entities,and the <cause events,causal relationships,result events> triples were constructed.Next,the vector space model is used to calculate the similarity between events to construct a triple of <event,similarity,event>;finally,the event is treated as an entity,the causal relationship between the event and the event,and the event and event are similar.Relationships and events and entity connections associated with events are added to the knowledge graph as relationships,further enriching and refining knowledge of knowledge graphs.Based on the final constructed financial knowledge graph,combined with the examples of “Changchun Longevity Vaccine Event” and “Chile Earthquake Event”,the application of knowledge graph in financial risk control and auxiliary decision analysis is analyzed,and the knowledge graph is analyzed.The advantages and value of these application scenarios.