K-SENSE integrates COMET-derived commonsense knowledge with parallel semantic encoding and contrastive training to reach F1 scores of 86.1 on stress detection and 94.3 on depression detection, beating prior baselines by 2.6 and 1.5 points.
MentalBERT: Publicly available pretrained language models for mental healthcare
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
K-SENSE: A Knowledge-Guided Self-Augmented Encoder for Neuro-Semantic Evaluation of Mental Health Conditions on Social Media
K-SENSE integrates COMET-derived commonsense knowledge with parallel semantic encoding and contrastive training to reach F1 scores of 86.1 on stress detection and 94.3 on depression detection, beating prior baselines by 2.6 and 1.5 points.