About The Project
Empowering Mental Health with NLP.
VibeCheck is a mental health text classifier built for students. Paste how you're feeling and the model identifies your emotional state across 7 categories — from everyday stress to clinical signals like suicidal ideation. The UI transforms its entire visual identity to match the result, making the experience feel as human as the problem it addresses.
The Models
Quick Vibe
MentalBERT Flat v3
A fine-tuned MentalBERT model (BERT-base-uncased with mental health pretraining) trained on a 7-class flat classification task. Fast inference — a single forward pass produces the result. Best for quick reads when you want a near-instant vibe.
81.98%
Accuracy
746
Dep→Sui errors
478
Sui→Dep errors
Deep Dive
Two-Branch + Longformer Stage 3
A 4-stage pipeline that routes text through a suicidal gate (Stage 1A), a normal/distress splitter (Stage 1B), a 5-class distress classifier (Stage 2), and finally a Longformer re-scorer that resolves ambiguous Depression vs. Suicidal calls using up to 1,024 tokens of context. Takes longer on first load, but significantly sharper on the cases that matter most.
86.97%
Accuracy
740
Dep→Sui errors
65
Sui→Dep errors
Dep→Sui: Depression samples misclassified as Suicidal. Sui→Dep: Suicidal samples misclassified as Depression — the more critical error. Deep Dive reduces this by 86%.
The Team
Project Supervision
Dr. Lamees Nasser
Assistant Professor
Eng. Mirna Ahmed
Teaching Assistant
Disclaimer
VibeCheck is an academic NLP project developed as part of a university course. It isn't a substitute for professional mental health support, clinical assessment, or any form of medical advice. All classifications produced by this system are model predictions based on input text and should not be interpreted as diagnoses of any mental health condition. If you or someone you know is struggling, please consult a qualified mental health professional.