On Reddit posts labeled by subreddit membership, transformer models, led by RoBERTa, reach 99.5% F1, far above LSTM baselines, but the proxy labels may inflate the apparent detection ability.
A review of machine learning and deep learning approaches on mental health diagnosis,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
support 1representative citing papers
citing papers explorer
-
Advancing Mental Disorder Detection: A Comparative Evaluation of Transformer and LSTM Architectures on Social Media
On Reddit posts labeled by subreddit membership, transformer models, led by RoBERTa, reach 99.5% F1, far above LSTM baselines, but the proxy labels may inflate the apparent detection ability.