Dissecting Twitter Discussion Threads with Topic-Aware Network Visualization

Twitter is the archetype of social networks designed for information exchange and discussion among users with similar interests and passions. Twitter prompts users to disseminate a new topic and get involved in forming opinions around it. It is always of interest how and why some tweets get more attention and involve more users, especially in the case of social issues and political debates. In this paper, we use an attention-based deep learning model to extract the topics of discussion on Twitter. Then, we reconstruct the network of discussion threads to visualize a topic-aware network of discussions. The proposed method yields the amount of attention given by users to different topics over time. As a case study, we applied our analyses to discussion threads around tweets by President Donald Trump during 9 months. Such analyses help us to understand how much each topic triggers discussions, how the topics are interrelated, and how opinions form around the tweets.