This article presents the first comprehensive review of non-social media, free-text datasets used for mental health research, aiming to address the sampling biases and ethical issues associated with social media data. Using the PRISMA methodology, the authors surveyed available datasets across multiple languages to evaluate their characteristics and utility.
- Non-social media free-text datasets are predominantly focused on the English language.
- The majority of these datasets target the detection of depression.
- There is significant variation in demographics, platforms, data types, annotation techniques, and methodologies among the reviewed resources.
- The review identifies key gaps in current resources.
The authors highlight opportunities to develop more diverse, reliable, and clinically relevant resources to overcome existing limitations in mental health disorder detection.