[3J10-14] Social Research & Communication 4

3:15 PM - 3:30 PM JST(6:15 AM - 6:30 AM UTC)

[3J12] Analyzing Reports on Nuclear Energy in Japan Through Topic and Sentiment Analysis

*Yifan Sun¹, Tsuruta Hirofumi², Kumagai Masaya^1,2, Ken Kurosaki¹ (1. Kyoto Univ., 2. SAKURA internet Inc.)

Keywords:

Atomic energy,Online media,Topic modeling,Sentiment analysis,Language models

In the aftermath of the Fukushima Daiichi nuclear power plant accident, understanding public sentiment and fostering effective communication is crucial for the future of nuclear power in Japan. This study leverages online comments on YouTube videos to gain deeper insights into public sentiment, which are often more expressive compared to traditional surveys with fixed responses. We extracted data from over 3,000 YouTube videos posted by official Japanese news channels up to April 1, 2024. Using a combination of the bag-of-words approach and Latent Dirichlet Allocation (LDA), we identified eleven major topics in these videos. Subsequently, we conducted sentiment analysis on the comments related to each of these topics, utilizing both lexicons and large language models, to evaluate viewer responses.