Research On Academic Knowledge Graph Based On Text Classification And Implementation Of Discipline Correlation Map

Since the release of relevant documents on "first-class disciplines" in China,major universities across the country have been actively promoting the construction of "first-class disciplines".The construction of "first-class disciplines" needs the support of discipline services.In the discipline services,such as discipline competitiveness analysis,discipline frontier analysis,discipline situation analysis,etc.,there are certain requirements for the discipline classification of documents.Mainstream academic knowledge base usually does not support the classification of documents according to the first-level disciplines of the Ministry of Education.Therefore,realizing automatic discipline classification of documents and adding support for disciplines of the Ministry of Education to the academic knowledge base are of positive significance to the construction of "firstclass disciplines".As a new way of realizing academic knowledge base,knowledge graph stores academic knowledge in the form of semantic network,provides highly scalable storage of academic data,and can efficiently support the first-level disciplines of the Ministry of Education by adding discipline entities.In addition,the academic knowledge graph has good support for data mining tasks.In addition,the academic knowledge graph has good support for data mining tasks,and the interdisciplinary colleration mining task can fully demonstrate the data mining ability and discipline support ability of the academic knowledge graph.In order to study the academic knowledge Graph based on text classification and realize the visualization of discipline correlation map,,the main work completed is as follows:This paper proposes an automatic discipline classification method of document based on the pre-training language model.The method uses the text classification technology to realize the automatic discipline classification of document,and realizes the support of academic knowledge graph to the first-level discipline of the Ministry of Education,which solves the problems of time-consuming,laborious,and large errors caused by the large workload of manual discipline classification of document and the limitation of personal knowledge range.A disciplinary correlation model based on internal and external correlation is proposed..The model uses the proportion of disciplines covered in the document,support and confidence to realize the calculation of multi-disciplinary correlation,which solves the problem that the cosine similarity algorithm cannot complete the calculation of three or more disciplines because it only calculates the angle between vectors,and provides good support for academic research such as multi-disciplinary integration;Design and implement the disciplinary correlation map visualization system.The system obtains academic data through web crawlers,uses the automatic discipline classification model of document to classify the document data,constructs the academic knowledge graph,uses the disciplinary correlation model to calculate the multidisciplinary correlation,and presents the results in the form of the disciplinary correlation map,and displays some additional correlation information.