Login to discuss
No comments here
Why not start the discussion?
In this paper, we explore topic modeling for the assessment of risk for depression, anorexia and self-harm. Using social media textual content from different datasets, we focus on Latent Dirichlet Allocation models, trained on both specific and combined corpora made from these datasets to perform risk detection. We investigate mental health vocabulary and shared topic modeling performance improvements on user classification.
Article ID: 2021S20
Venue: Canadian Conference on Artificial Intelligence
Publisher: Canadian Artificial Intelligence Association