Earticle

초록

영어

With the proliferation of generative AI painting technology, “AI painting” has sparked widespread discussion on social media. The Weibo platform, with its large user base and high interactivity, provides an ideal context for exploring public attitudes and topic dynamics. This study combines sentiment analysis and topic modeling to quantitatively analyze 13,379 relevant Weibo posts published between October 2024 and March 2025. For sentiment analysis, we constructed a BERT-based model incorporating a dynamic attention mechanism. After pre-training and fine-tuning on 1,487 domain-specific texts, the model achieved an accuracy of 95.12% and an ROC-AUC of 0.9818. The results reveal a polarized sentiment among Weibo users toward AI painting: 75.9% expressed positive emotions, while 24.1% expressed negative emotions. Positive content tended to combine images and text and received more reposts, whereas negative content was more often original text-only posts but demonstrated a greater capacity to drive interaction, reflecting the "negativity bias" phenomenon. In terms of topic modeling, BERTopic identified five core themes: ethical controversies and industrial impact, technology-driven creation, artistic exploration and aesthetic innovation, digital preservation of cultural heritage, and vertical content operations. These themes highlight the multifaceted significance of AI painting across technological, artistic, and social dimensions. From the perspectives of communication studies and psychology, this study uncovers the emotional structure and communication mechanisms of “AI painting” on Weibo, offering empirical and methodological insights for understanding the social acceptance of generative AI and platform public opinion governance.

목차

Abstract
Introduction
Theoretical Background
Online buzz
Social Media
Generative Technology
Methodology
Data Collection and Preprocessing
Sentiment Analysis Modeling
Topic Analysis Modeling
Results
Sentiment Analysis
Topic Analysis
Conclusion and Discussion
References

저자

• Taekyung Kim [ Big Data Analytics Kyung Hee University, Professor ]
• Li Zhaoying [ Big Data Analytics Kyung Hee University, MS ]

참고문헌