Empirical Research of Hot Topic Recognition and its Evolution Path Method for Scientific and Technological Literature

抄録

With the advent of big data era, the recognition of hot topics and the analysis of their evolution path in the frontier of a certain field of scientific and technological literature have received widespread attention from the academic community. It can not only reveal the development trend in a certain field of scientific and technological literature, 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, this paper uses Latent Dirichlet Allocation (LDA) model to propose a hot topic recognition and evolution analysis method in scientific and technological literature field, 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. Main research process is as follows. Firstly, LDA is used to extract global topics and stage topics. Secondly, similarity calculation algorithm is used to filter topics. Thirdly, novelty and support are used to identify hot topics. Fourthly, three paths of inheritance evolution, merging evolution and division evolution are formed for hot topics. Finally, the effectiveness of the method is verified by using 47,896 scientific and technological literature data in the field of intelligent algorithms in Web of Science as an empirical example.