Covid-19 Tweets Analysis with Topic Modeling
Shichao Jia, Qi Chen, Wei Wang · 2021
Social media has become an important data resource for knowledge discovery and data mining in multiple disciplines. With the exploding amount of social media data, how to efficiently and effectively exploit values and insights from such overwhelming amount of data has become an emerging area. Recently, various natural language processing techniques have been developed, e.g., word embedding, deep neural network and Latent Dirichlet Allocation (LDA), for studies such as sentiment analysis, traffic event detection, nature disaster assessment and COVID-19 tweet analysis. In this paper, topic modeling through LDA was used to conduct text mining on a large real-world COVID-19 tweet dataset, which contains more than 524 million multilingual tweets and covers 218 countries over a period of 3 months. We conducted extensive experiments and visualise insights discovered through this unsupervised process.