Time-series Analysis of Newspaper Articles for Automatic Event Detection using LDA
Watanabe Arisa, Basabi Chakraborty · 2021
Newspapers are useful for understanding important past events. However, reading all articles is difficult and time-consuming. Automatic detection of characteristic events from past time series text data helps an interested user to find important past events. Automatic topic classification of newspaper articles and visualizing the change of topics over a certain period are used for past event detection. In this study, automatic topic classification is performed using Latent Dirichlet Allocation (LDA). The daily changes in the topics are measured by using different distance metrics to detect the changes which indicate important events. The idea has been implemented with a simple and small data set to check its validity.