Machine Learning-Based Data Mining Method for Sentiment Analysis of the Sewol Ferry Disaster's Effect on Social...

Area of Science:

• Social Sciences
• Computational Social Science
• Natural Language Processing (NLP)

Background:

• The 2014 Sewol Ferry Disaster inflicted widespread social distress across South Korea.
• Previous domestic research has not quantitatively assessed the disaster's impact on social stress using social media sentiment analysis.

Purpose of the Study:

• To analyze public sentiment on social media platforms (YouTube, Twitter, Facebook) following the Sewol Ferry Disaster.
• To investigate the influence of the disaster on social stress through sentiment analysis of user-generated content.

Main Methods:

• Collected and analyzed text data from YouTube, Twitter, and Facebook users who posted about the disaster (April 2014 - March 2015).
• Employed sentiment analysis, including NLP-based data mining (phrase, entity, and query analysis), word clouds, and bar graphs.
• Utilized ANOVA for statistical comparison of sentiments at a 95% confidence level.

Main Results:

• A significantly negative sentiment was observed across all analyzed social media platforms.
• Negative sentiment was strongly associated with political entities, including ex-president Park and associated politicians.
• Negative sentiment prevalence was high in phrases (58.9-69.4%), entities (69.9-81.1%), and query topics (75.0-85.4%) across platforms.

Conclusions:

• This study provides the first scientific evidence of the negative psychological impact of the Sewol Ferry Disaster on the Korean population via social media.
• The findings underscore the significant role of social media in reflecting and potentially amplifying public distress during national crises.