Research On Extraction Of Evolution Event Knowledge Of Organization For The Construction Of Organization Knowledge Graph

With the rapid development and continuous evolution of computer and Internet technology,the demand for intelligent and knowledge-based services in the society has become increasingly obvious,thus giving rise to more research on construction and application of knowledge graph.Among the knowledge graphs in many fields,more and more attention has been paid to the research on construction and application of knowledge graph in organization's history and culture field.On the one hand,organization knowledge graph will provide support for the research on the history of science and technology and the communication of scientific culture.On the other hand,organization knowledge graph is an important tool for intelligent information analysis.The evolution event knowledge of organization,which describes the change process of the organization and the relationship between organizations,can be combined with important figures,important achievements and major events to form organization knowledge graph in order to meet the demand of knowledge support for intelligent information analysis and communication of science and technology history and culture.Based on the research of science,this study constructs a model of the evolution event knowledge of organization according to the characteristics of the development and evolution of scientific organization.It also puts forward a technical scheme of extracting evolution event knowledge of organization.The technical scheme is verified and improved in the experiment,which can provide reference for related research.Specifically,this study mainly completed the following work.First of all,this study makes an extensive research on extraction techniques of knowledge such as entity and relation,models of event knowledge and extraction techniques of event knowledge.And it summarizes the advantages and disadvantages of various models and algorithms as well as their applicability in the field.Secondly,this study constructs a conceptual model of the knowledge according to the relevant models of event knowledge and the relevant ontology at home and abroad.And it also designs a representation with quaternion according to the requirements of actual storage and application.Then,based on the relevant technology and characteristic of the evolution event knowledge model of organization,this study proposes a pipeline extraction technical scheme of evolution event knowledge of organization,which includes the following contents.(1)This study proposes an event detection method based on BERT-WWM to filter the irrelevant sentences.(2)Then,it puts forward to an organization identification method based on semi-automatic corpus and deep learning method,which combines the fusion method of lexicon and feature with bidirectional long-short memory neural network and conditional random field.At the same time,it uses Standford NER to identify event time.(3)Referring to the event classification with role determination,this study proposes a trigger recognition and event classification method based on trigger words list extension,and then comes up with a recognition model of computing convolution based on improved syntax tree(dparse-patt-cparse?CTKM)to annotate the role of organization,link the event time and combine the quaternion.After role annotation,event type element is improved with role annotation results.An experimental platform is built for the extraction experiment of evolution event knowledge.And the results of the experiment showed that the technical scheme achieves valid and the F1 score was about 64.37%.The results of the experiment basically meet the demand of extracting the evolution event knowledge of organization from unstructured Chinese historical texts.The experimental results of each stage prove that each method is effective and can provide new technical solutions for entity extraction,event role judgment and text classification.This dissertation consists of 25 pictures and 23 tables.