ABSTRACT
By 2019 COVID-19, since the epidemic, the number of relevant documents exponentially level rise. Faced with a large amount of literature, this research provides convenience for exploring the connection between research topics and fields and quickly understanding relevant literature information. We pass on the data set after data cleansing using the LDA(Latent Dirichlet allocation) methods, and Berts and K-means modeling method extracting topic keywords. Use knowledge graph tools to output relevant visual graphics and systematically extract adequate information. Through text mining of biomedical research papers related to COVID-19, the improved model is used to analyze and make recommendations to respond to and prevent the COVID-19 pandemic. This research can support the rapid and in-depth analysis of a large number of relevant documents and can be used in future research to support real-time scientific disease research. © 2022 IEEE.