Automatic detection of eating disorder-related social media posts that could benefit from a mental health intervention.

ABSTRACT:

Objective

Online forums allow people to semi-anonymously discuss their struggles, often leading to greater honesty. This characteristic makes forums valuable for identifying users in need of immediate help from mental health professionals. Because it would be impractical to manually review every post on a forum to identify users in need of urgent help, there may be value to developing algorithms for automatically detecting posts reflecting a heightened risk of imminent plans to engage in disordered behaviors.

Method

Five natural language processing techniques (tools to perform computational text analysis) were used on a data set of 4,812 posts obtained from six eating disorder-related subreddits. Two licensed clinical psychologists labeled 53 of these posts, deciding whether

SUBMITTER: Yan H 

PROVIDER: S-EPMC6790146 | biostudies-literature | 2019 Oct

REPOSITORIES: biostudies-literature