Research On Knowledge Combination And Knowledge Dissemination In Cross-discipline Based On Latent Topics

In the development of modem science,the division of science is always accompanied by emerging scientific research activities which break the boundaries of traditional disciplines or research fields and gradually develop into cross-disciplines.Interdisciplinary research has become one of the main directions of modern scientific development,belonging to the forefront of science.In this context,the study of interdisciplinary knowledge innovation and operational mechanism will help to further promote the prosperity and development of cross-disciplinary research.In essence,the knowledge innovation in cross-disciplines is the result of the input,composition and integration of knowledge from relevant basic disciplines.The research on knowledge structure and knowledge communication of cross-disciplines can benefit the analysis on operation mechanism of cross-disciplines from the source of innovation and reveal the development law of cross-disciplines.This study identifies latent topics from the content of literature on cross-disciplines.In the point of view of topics,this study identifies the relationships among topics on cross-disciplines with the purpose of disclosing the structure and laws under knowledge integration and knowledge dissemination of the cross-disciplines to spy on the running mechanisms of cross-disciplines.This paper studies knowledge composition and knowledge dissemination in cross-disciplines from the cut-off perspective of latent topics.The paper first proposes a method to construct an integrated dataset consisting of literature on the cross-discipline and basic disciplines.Latent topics are identified from the integrated dataset.Then,two analytical routes are applied in this study.One is the co-occurrence relationship between topics.It's assumed that the co-occurrence among topics in scientific literature is a reflection of the combination of knowledge.Based on this hypothesis,topic co-occurrence network is constructed to study knowledge composition structure and patterns of interdisciplinary knowledge combination in cross-disciplines.The second relationship is citation relationship,which symbolize the spread of scientific knowledge among topics.Topic citation network is constructed to study the structure of knowledge dissemination.In the first chapter,basic theories related to interdisciplinary research are sorted out,including the concept of interdisciplinary and interdisciplinary measurement,and the theory and methods on topic,which are fundamental for this study.In the second chapter,we propose a method to identify topics from the integrated data sets of literature from cross-disciplines and related basic disciplines.The underlying hypothesis of this method is that interdisciplinary research is based on relevant basic disciplines,so identifying the topics should not only based on the literature from the cross-discipline but from related basic disciplines.Only doing so,it's more accurate to identify topics from interdisciplinary research.This paper combines expert identification and bibliometric analysis to identify related basic disciplines for cross-disciplines,and then constructs the corresponding integrated dataset.Latent Dirichlet Distribution model is applied to identify topics from the integrated dataset.In addition,latent topics of the articles are extracted as well,so that the number of documents per topic is obtained.In the third chapter and the fourth chapter,this article holds that co-occurrence of topics in cross-disciplines is the embodiment of knowledge combination in the process of knowledge innovation in interdisciplinary research.First,we construct topic co-occurrence network,which is based on the co-occurrence relationships between topics in the cross-discipline.The nodes are the topics in the cross-discipline and the edges form for topic pairs that co-occur in papers with weights representing the number of papers they appear.Then,social network analysis and graph mining algorithms are applied to analyzing topic co-occurrence network.Major methods are as follows:First,node centrality metrics and random walk algorithm is used to measure the importance of different research topics and the roles of topics are explained correspondingly;Second,the weight of edges are analyzed to identify frequently-co-occurred topic pairs;Thirdly,community detection algorithm is applied to identify topic community,and analyzing the phenomenon of knowledge cluster among topics.Then,this paper will build multi-mode topic network by considering the attributes of topics.The attributes of topic to be considered include:?topic type.The topics are divided into two types of research objects and research methods;?topic disciplines.According to how articles are distributed into the topics of cross-discipline and related basic disciplines,the disciplines that topics belong to are determined by combining with expert experience.Patterns of interdisciplinary knowledge combination in cross-disciplines are disclosed by analyzing multi-mode topic network.On one hand,the patterns on discipline combination are quantified by doing statistic computation on papers and topics.On the other hand,the combination patterns of research objects and research methods can be discovered by analyzing nodes and edges.The fifth chapter investigates the knowledge dissemination structure of cross-discipline based on topic citation network.First,the paper citation networks in cross-discipline and in the integrated dataset are transferred into topic citation networks with directed weighted edges.In cross-discipline,topic cited frequency and topic cited diversity are indexes that measure the impact of topics.Two levels of impact are investigated.One is the impact to other topics inside the cross-discipline and the other is the impact from the integrated dataset to other topics in the cross-discipline.Finally,topic-level knowledge dissemination map is constructed by considering the impact of topics,which can be used to reveal knowledge dissemination structure within cross-discipline and in the integrated dataset.In this study,digital library is selected as an example cross-discipline for empirical study.The results show that LDA is suitable for identifying latent topics for digital library discipline.The knowledge composition and knowledge dissemination analysis framework in the perspective of topics is a successful tool to analyze this cross-discipline in a fine-grained level.This framework is based on topic occurrence network,discipline-subject-method topic network and topic citation network.At the end,this paper summarizes the whole work and puts forward the shortcomings of this study and the future research.