Spatiotemporal Analysis of Electronic Cigarette Perception on Twitter/X Using Natural Language Processing.

Area of Science:

• Social Media Analytics
• Public Health Surveillance
• Digital Communication

Background:

• Electronic cigarettes (e-cigarettes) have surged in popularity, particularly among young demographics.
• Understanding public perception of e-cigarettes is crucial for public health initiatives and policy development.

Purpose of the Study:

• To analyze the spatiotemporal patterns of online perception regarding e-cigarettes on Twitter/X.
• To identify key themes and sentiments expressed in e-cigarette-related social media discussions.

Main Methods:

• Collected over 3 million e-cigarette-related tweets from March 2021 to March 2023 via the Twitter API.
• Utilized Natural Language Processing (NLP) techniques, including RoBERTa and Latent Dirichlet Allocation (LDA), for sentiment analysis and topic modeling.
• Human coders validated tweet relevance, sentiment, and user type for model training and analysis.

Main Results:

• Observed a significant increase in e-cigarette discussions, notably in the UK and Australia.
• Positive sentiment (27.0%) slightly exceeded negative sentiment (23.3%), with neutral tweets comprising nearly half.
• E-cigarette users expressed significantly more positive sentiment (41.19%) compared to non-users (9.74%).

Conclusions:

• Online perceptions of e-cigarettes on Twitter/X demonstrate temporal and geographical variations.
• Distinct sentiment differences exist between e-cigarette users and non-users on social media.
• Findings offer valuable insights for targeted public health messaging and evolving tobacco regulations.