Identifying E-cigarette Content on TikTok: Using a BERTopic Modeling Approach.

Area of Science:

• Social Media Analysis
• Computational Linguistics
• Public Health

Background:

• Social media platforms like TikTok are increasingly used to promote e-cigarette products.
• Analyzing vast amounts of social media data for content promotion is challenging.
• Hashtag analysis offers insights into e-cigarette promotion strategies online.

Purpose of the Study:

• To apply machine learning for identifying e-cigarette content themes on TikTok.
• To analyze the nature of e-cigarette promotion and discussion on the platform.
• To evaluate the effectiveness of BERTopic modeling for social media content analysis.

Main Methods:

• Utilized 13 unique e-cigarette-related hashtags for data collection on TikTok.
• Analyzed a dataset of 12,573 TikTok posts using Bidirectional Encoder Representations from Transformers (BERT) topic modeling.
• Employed quantitative (coherence test) and qualitative (relevance checking) methods to determine the optimal number of topic clusters (N=18).

Main Results:

• Identified 9 overarching themes within the e-cigarette TikTok content.
• Key themes included social media features, vape shops/brands, vape tricks, modified e-cigarette use, vaping and gender, flavors, association with traditional cigarettes, and community aspects.
• Discovered non-English language content, specifically Romanian and Spanish, indicating a global reach.

Conclusions:

• BERTopic modeling effectively identified relevant themes in e-cigarette content on TikTok.
• This machine learning approach can be a valuable tool for future social media research on tobacco products.
• Findings can inform tobacco regulatory policies, particularly for monitoring e-cigarette marketing on social media.