Exploring Trending Topics of Social Media Text with VoronoiTopicCloud Provide Useful and Intuitive Insights into Social Media Texts
Yuhui Wang, Feng Zhou, Xiaoyong Li · 2021
Enormous volume of data on social media platforms such as Weibo and Twitter is being generated every day. Microblogs collected from Weibo on a certain topic may consist of numerous conversation threads about relevant subtopics. However, it is difficult to distinguish these subtopics if the data is visualized as a single WordCloud because there’s no contextual information around the words in WordCloud. To help identify subtopics under a broader topic, we introduce VoronoiTopicCloud, a novel technique for visualizing the opinions in a large social media text collection. In order to overcome the problem of semantic sparseness and improve the performance of text clustering, we propose a novel corpus-based enrichment approach for short text. In the expansion procedure, new words which may not appear in the original short text document are added to the document with a virtual weight and the virtual weight is obtained from the posterior probabilities of new words given all the words in that document. At last, VoronoiTopicCloud is implemented as a lightweight application that runs in the browser.