Evolution of Topic Using LDA Model: Evidence From Information Science Journals

To learn about the trend and monitor the hot topic of research, mining evolution of topic intensity from papers plays an important role. A model based on Latent Dirichlet Allocation (LDA) was proposed. Firstly, the collections of all the papers by using LDA to find out topics and their key words, and probability distribution of documentation-topic on different time windows was obtained. Secondly, LDA was applied in papers on every single time window to get probability distribution of topic-word, through which similarity of topics from different time windows were computed. The trend of topic intensity was figured out, and the words probability of similar topics can help figure out the trend of topic content. It shows that the topic of semantic analysis draws more and more attention in the field of Chinese informatics.