Social Media Communities Topic Modeling

Abstract

Social media is a source of native information generated by user communities. Processing a large amount of text and the related information posted in social media is an urgent task for developers of modern intelligent information systems. Such systems are necessary for information support for decision-making in regional management. Topic modeling is a method of constructing text document collection models, it is used in solving problems of exploratory search, classification and annotation of document collections and news streams. The paper explores the application of different topic modeling methods on social media texts. The specifics of text content are discussed, as well as the methods of assessing the applicability of topic modeling methods in the aspect of social media communities modeling. For different topic modeling methods, the experiment results on calculating UMass metrics with a varying number of topics on datasets created by the authors are presented. Conclusions are made about the preference of the BigARTM library for further work on modeling social media communities.