Text Classification of Sudden Public Health Events Based on Abstracts of Scientific Papers

In this study, text analysis techniques are used to explore the abstracts of papers in the field of public health emergencies, with the aim of accurately capturing the research dynamics and directions in this field and providing rich references for researchers. In this paper, we first use Latent Dirichlet Allocation (LDA) to explore the 10 topics corresponding to the relevant literature, and classify the literature based on the topics. At the same time, considering the randomness of the LDA model parameters, this study incorporates Long Short-Term Memory (LSTM) to obtain a new hybrid model to improve the accuracy and stability of text classification, and finally the model obtains 15 topics and discusses the evolution of the research topics of the literature in this field in the past 20 years, which still has great prospects for development in the future.