01/19/2021
By Dedra Daigle
The Department of Public Health presents "Mining Social Media to Detect Alcohol Use Risk at Population Scale" on Friday, Jan. 22 from 2 to 3 p.m. The Zoom session is presented by Hadi Amiri. Please contact Dedra Daigle@uml.edu for Zoom link if interested.
User Generated Content, which can range from social media discussions to product reviews to private physician notes, present naturally occurring data that can be used to develop large-scale machine learning algorithms to obtain low-cost and high-resolution views into population behavior, in particular, the unhealthy behaviors such as hate speech, sexual harassment, fighting, alcohol abuse, etc. In this talk, I will discuss the development of an effective online surveillance system to detect alcohol use risk at population scale. Alcohol use is associated with leading causes of morbidity and mortality, and the high magnitude and cost of its related problems make it especially vital to identify opportunities for preventive interventions. I will discuss the problem of identifying first-person reports of alcohol drinking, consumption level (light or heavy) and drinking context (individual or group drinking) from social media feeds for population-level analysis. Our model utilizes extensive linguistic features (including word dependencies and language bias features) and obtains high accuracy in detecting the above indicators from alcohol-relevant posts. Such data could provide complementary information for monitoring alcohol use, informing prevention and policy evaluation, and extending knowledge available from national surveys.