Research On Modeling And Evolution Mechanism Of Discipline Knowledge Network Based On Integrating Topic

With the development of science and technology,scientific research has entered into the fourth paradigm of data-intensive scientific discovery.Under the new knowledge ecology,the users oriented to discipline knowledge innovation put forward higher requirements for knowledge service.However,the traditional knowledge service exists some imperfects,such as,not enough in-depth for revealing knowledge semantic content,not enough strong links between the knowledge points of relevance,not enough fine for describing the development and evolution regularity of the discipline and so on.Based on this,the research objectives of this article include three respects.In respect of theoretical research,this article explored formation and evolution of knowledge network.In respect of method research,this article realized deep mining and evolutionary mechanism analysis of latent discipline knowledge structure.In respect of applied research,this article provided analytical predictive knowledge services supported bydiscipline knowledge network based on integrating topic.The basic research idea of this article isalong the way of theoretical research,network building,evolution mechanismand modeling simulation.The main work and innovation are divided into the following several parts.Part one focused on basic theories of discipline knowledge network.Firstly,the paper systematically analyzed the basic theory of three related disciplines,which warescience of science,text mining and network science.Secondly,the paper definedconcepts of discipline knowledge and discipline knowledge network,so as the connotation,types and basic 'characteristics.Then,the dynamic social network modeling method was introduced to construct the discipline knowledge network.Four different elements of scientists,knowledge unit,literature and academic institutionsware used to build a variety of types of discipline knowledge network through complex relationships between them.Finally,in aspect of network evolution mechanism,the basic driving processes of network changing ware analyzed,and the evolution mechanisms behind the related social theories ware discussed.Part two explored the theories and methods of discipline topic miningin great depth,from the view of topic representation and model optimization.Firstly,the paper cleared the meaningof topic and selected themethod of topic extraction which was the best for buildingdiscipline knowledge network.The results showed that topic extraction based on probability topic models was the best choice,such as LDA topic model.Secondly,the LDA topic model was improved and optimized from aspects of determining the best topic number,corpus building and effect evaluation of topic extraction.Finally,the article innovatively putted forward an evolution analysis framework of topic mining of authors,based on life cycle theory and the AT topic model.The results showed that the methods played an important supportingrole in revealing development process of discipline.Part three systematically studied construction and evolution analysis of discipline knowledge networkbased on integrating topic,from the view of correlation calculation between topic and other knowledge nodes.Three types of network ware analyzed in detail from one mode network structure,two modes network structure and collaboration network.Firstly,in the respect of one mode network study,the concept of topic network influence was proposed for finding key topics in the network.Through network structure analysis,this network had the typical characteristics of small world network.In the respect of two modes network study,the paper didn't take the projection method and retained the original information of the authors and topics.The paper putted forward measurement method of author topic correlation influence.The empirical study showed that scholars with high degree of research activity,broad research scope and potential cooperation space could be found by this influence index.Finally,discipline knowledge collaboration network based on co-authorship and relationship of owning same research topics was built,in the purpose of analyzing cooperation evolution mechanism.Part fourputted forward the evolution mechanism and model of discipline knowledge collaboration network,from the perspective of topic similarity.Firstly,the paper proposed that research topic similarity between scholarswas the important foundation of scientific cooperation through empirical analysis.Secondly,the discipline collaboration network evolution model(RTBA for short)wasinnovatively putted forward based on node connection mechanism of research topicsimilarity.The simulation experiment results showed the feasibility and validity of this model.Finally,based onRTBAmodel and NWS model,the knowledge diffusion phenomenon in scientific collaboration networkwas studied.This paper considered that knowledge diffusion in scientific cooperation network wasmainlyinfluenced by two processes,which wareknowledge spillover and knowledge innovation.The paper found main conclusions as follows.Knowledge spillover mainly influenced knowledge diffusion in the early period of knowledge diffusion,whichwas more sensitive to different network structure.Individual knowledge innovation mainly influenced knowledge diffusion in the later period,whichwas not sensitive to different network structure.