[1L4-GS-5] Web intelligence: Social data analysis (1)

[1L4-GS-5-01] Twitter Analysis of 2019 Osaka Double Election

〇Eiichi umehara¹, Yuuki Ogawa², Atuki Hirakawa¹ (1. Tokyo City University, 2. Ritsumeikan University)

Keywords:Osaka Double Election, Social Network Analysis, Natural language processing, Announcement effect

We analyzed the announcement effect and selective exposure of Twitter posted during the Osaka Double Election in 2019 using topic analysis and social network analysis. First, we used LDA to categorize tweets into 10 topics, and categorized each topic into supporters and opponents of One Osaka. Supporters and opponents had almost the same in the early stages, but in the late stage, supporters were dominant. This might be due to a newspaper report from the opinion poll on April 1 stating that supporters were dominant. As a result, the supporters might have a bandwagon effect and the opposition might have an underdog effect. Second, we conducted a social network analysis on retweets. Using the Louvain method, large communities were extracted in each stage. Within each community, there was no supporter account with high centrality and opposition together. As a result, many users might have made selective exposure in Twitter.