Abstract
This study examines the effectiveness of various sentiment analysis techniques on Douyin, specifically focusing on fashion-related content in the Chinese context. As Douyin continues to dominate the social media landscape, particularly among Generation Z in China, understanding sentiment on this platform is essential for brands and businesses seeking to connect with this digitally savvy demographic. The research compares four sentiment analysis methods—Support Vector Machines (SVM), VADER (Valence Aware Dictionary for sEntiment Reasoning), Convolutional Neural Networks (CNN), and Bi-directional Long Short-Term Memory networks (biLSTMs)—to analyze user-generated content on Douyin. Data for the study were collected through scraping videos related to prominent fashion brands, followed by manual labeling to facilitate sentiment analysis. The results demonstrate varying levels of accuracy across the methods, with each showing unique strengths and limitations depending on the informal and evolving nature of Douyin content. While no single method emerged as consistently superior, the findings highlight the challenges of applying sentiment analysis to dynamic social media platforms and suggest the need for hybrid or context-aware approaches. This research contributes to the growing field of sentiment analysis in the Chinese digital ecosystem and offers valuable insights for adapting sentiment analysis tools to local linguistic and cultural nuances. It also underscores the importance of refining analysis techniques to account for the rapid evolution of online language and platform-specific communication styles. The study provides implications for marketers and researchers seeking to design more effective sentiment analysis strategies tailored to the preferences and behaviors of Chinese Generation Z users.
Presenters
Shurui ZhengStudent, Broadcasting and Hosting Art, Yunnan Arts University Wenhua College, China
Details
Presentation Type
Paper Presentation in a Themed Session
Theme
KEYWORDS
Sentiment Analysis, TikTok, Generation Z, Natural Language Processing
