Topic Discovery And Evolution Analysis Based On Literature Association Structure

With the development of global informatization,the massive growth of various digital documents has made it difficult to acquire knowledge.The literature published in the academic field presents the phenomenon of subject refinement and cross-integration.Therefore,how to analyze the distribution of topics from the massive literature and mining the evolution trend of topics over time is of great significance for analyzing the knowledge base,gaining insight into the hotspots in the field,and formulating macro strategic plans.At present,the two types of academic topic discovery research systems based on citation association and semantic association technology have each been well developed,but the methods and applications of integrating different association technologies need to be further explored;in addition,the research on the topic evolution analysis mainly focuses on The linear evolution of a single topic,while the nonlinear evolution research that can identify multiple types of evolution events between topics is not yet mature.Aiming at the above situation,this paper researches academic topic discovery and topic evolution analysis based on document association structure.First,the paper proposes the MVSC-CTA algorithm and LPSC-T algorithm based on spectral clustering.By fusing the effective information of document coupling and document semantic coupling association structure according to different strategies,the complementation of citation and semantic information is realized.The experimental results show that the proposed two proposed algorithms both have good results in academic topic discovery.Secondly,the paper designs a nonlinear topic evolution framework based on pre-time discreteness and conducts empirical research on the topic evolution events based on the literature in the field of the Internet of Things.It can further realize the division of academic sub-fields and the extraction of different theme evolution paths.Finally,according to the research of the above algorithm and evolution framework,the thesis designs and implements an academic topic evolution analysis system based on the multi-relational structure of documents.Starting from the actual needs of researchers,the system can efficiently,stably,and parallel performs subject evolution analysis tasks and visualize the analysis results accordingly.