[4C1-GS-11] AI and Society:

[4C1-GS-11-01] Evolution of COVID-19 Vaccination-related Topics on Twitter

Hypothesis Building and its Verification based on Word Frequency Analysis and Close Reading of Tweets

〇Yuka Takedomi¹, Yuri Nakayama², Towa Suda¹, Takeaki Uno¹, Takako Hashimoto³, Masashi Toyoda⁴, Naoki Yoshinaga⁴, Masaru Kitsuregawa⁴, Ryota Kobayashi^2,5 (1. National Institute of Informatics, Tokyo, 2. Graduate School of Frontier Sciences, The University of Tokyo, 3. Chiba University of Commerce, 4. Institute of Industrial Science, the University of Tokyo, 5. Mathematics and Informatics Center, the University of Tokyo)

Keywords:social media analysis, Covid-19, vaccine, Twitter, narrative

How have people's interests and public opinion evolved during the COVID-19 vaccine campaign? To address this question, we collected and analyzed all the Japanese tweets containing the word "vaccine" (“waku-chin”) from January 1st to October 31st, 2021. First, we used Latent Dirichlet Allocation (LDA) to perform automated topic modeling. Second, we manually categorized these 15 topics into the following four themes: (1) personal issue, (2) breaking news, (3) politics, and (4) conspiracy and humor, in order to gain a more comprehensive understanding of the discourse. Third, we constructed hypotheses about the evolution of topics by interpreting the narratives underlying the tweets through close readings of approximately 15,000 representative tweets. Finally, we verified the hypotheses by analyzing changes in the frequency of keywords. We found that tweets about personal issues have changed according to the actual situation of the vaccination campaign. Tweets expressing fear or strong opinions were more common during periods of greatest uncertainty about vaccination prospects and decreased as the situation stabilized.