Detecting Binge Drinking and Alcohol-Related Risky Behaviours from Twitter's Users: An Exploratory Content- and...

Area of Science:

• Social Science
• Public Health
• Computational Linguistics

Background:

• Binge drinking (BD) is a prevalent risky behavior often underreported to healthcare professionals.
• Social media platforms host numerous user-generated content related to alcohol consumption and behaviors.
• Detecting BD through online discourse presents an opportunity for public health surveillance.

Purpose of the Study:

• To develop and validate a data-driven method for identifying potential binge drinkers using social media content.
• To analyze linguistic features and topics associated with binge drinking behaviors online.
• To differentiate genuine user accounts from bots and commercial entities discussing alcohol-related topics.

Main Methods:

• Gathering Twitter data using expert-identified keywords related to binge drinking and alcohol behaviors.
• Manual labeling of a random tweet sample for training supervised learning classifiers.
• Utilizing linguistic and metadata features to classify user accounts (genuine, bot, commercial).
• Training a third classifier on genuine user data to detect potential binge drinkers based on language.

Main Results:

• Approximately 55% of 1 million alcohol-related tweets were identified as non-genuine user content.
• Potential binge drinkers identified through linguistic analysis showed similarities to general genuine users.
• The study successfully differentiated genuine users from bots and commercial accounts.

Conclusions:

• Social media user-generated content offers a promising avenue for identifying individuals engaging in risky behaviors like binge drinking.
• Linguistic analysis of genuine user posts can aid in understanding and potentially intervening in public health issues.
• This data-driven approach supports the development of informed preventive strategies for binge drinking.