RETRACTED ARTICLE: Research on Hot Topic Recognition and Its Evolution Analysis Method Based on LDA

Abstract

In recent years, the recognition of hot topics and the analysis of their evolution path in the frontier of a field have received widespread attention from the academic community. It can not only reveal the development trend of a certain field but also discover the evolution law of topic content in different development stages of the field. However, there are still some problems in some current research methods, such as inaccurate recognition of hot topics and unclear evolution path, which seriously affect the comprehensiveness and accuracy of the analysis. To solve the above problems, the paper uses LDA (Latent Dirichlet Allocation) model to propose a hot topic recognition and evolution analysis method, which aims to reveal the evolution law of topic content level in different development stages of the field, such as inheritance, merging, division, and other topic evolution trends so as to provide decision support for domain knowledge innovation services. LDA adopts the Gibbs symmetric sampling method, which can make the adopted data converge, and under the given conditions, the data is symmetrical in the high-dimensional space structure, which makes the analysis result more accurate. Main research process: Firstly, LDA is used to extract global topics and stage topics. Second, the topics are filtered using a similarity calculation algorithm. Thirdly, novelty and support are used to recognize hot topics. Finally, three paths of inheritance evolution, merging evolution, and division evolution are formed for hot topics.