|
|
--- |
|
|
language: en |
|
|
license: apache-2.0 |
|
|
library_name: transformers |
|
|
tags: |
|
|
- mental-health |
|
|
- text-classification |
|
|
- bert |
|
|
- nlp |
|
|
- depression |
|
|
- anxiety |
|
|
- suicidal |
|
|
datasets: |
|
|
- sai1908/Mental_Health_Condition_Classification |
|
|
- kamruzzaman-asif/reddit-mental-health-classification |
|
|
metrics: |
|
|
- accuracy |
|
|
- loss |
|
|
model-index: |
|
|
- name: bert-finetuned-mental-health |
|
|
results: |
|
|
- task: |
|
|
type: text-classification |
|
|
name: Text Classification |
|
|
dataset: |
|
|
name: sai1908/Mental_Health_Condition_Classification |
|
|
type: text |
|
|
metrics: |
|
|
- name: Accuracy |
|
|
type: accuracy |
|
|
value: 0.9656 |
|
|
- name: Validation Loss |
|
|
type: loss |
|
|
value: 0.1513 |
|
|
--- |
|
|
|
|
|
# BERT Fine-Tuned for Mental Health Classification |
|
|
|
|
|
This model is a fine-tuned `bert-base-uncased` transformer trained to classify text inputs into seven mental health categories. It is designed to support emotional analysis in mental health-related applications by detecting signs of psychological distress in user-generated content. |
|
|
|
|
|
## Try It Out |
|
|
|
|
|
You can interact with the model in real-time via this Streamlit-powered Hugging Face Space: |
|
|
👉 [**Live Demo on Hugging Face Spaces**](https://huggingface.co/spaces/Elite13/mental-health) |
|
|
|
|
|
## Datasets Used |
|
|
|
|
|
1. **sai1908/Mental_Health_Condition_Classification** |
|
|
Reddit posts from mental health forums |
|
|
~80,000 cleaned entries from the original 100,000 |
|
|
|
|
|
2. **kamruzzaman-asif/reddit-mental-health-classification** |
|
|
Additional Reddit mental health posts to improve coverage and diversity |
|
|
|
|
|
## Model Overview |
|
|
|
|
|
- **Base Model**: `bert-base-uncased` |
|
|
- **Type**: Multi-class text classification (7 labels) |
|
|
- **Framework**: Hugging Face Transformers |
|
|
- **Training Method**: Trainer API (PyTorch backend) |
|
|
|
|
|
## Target Labels |
|
|
|
|
|
- Anxiety |
|
|
- Depression |
|
|
- Bipolar |
|
|
- Normal |
|
|
- Personality Disorder |
|
|
- Stress |
|
|
- Suicidal |
|
|
|
|
|
## Training Configuration |
|
|
|
|
|
| Parameter | Value | |
|
|
|-----------------------|-------------------------| |
|
|
| Epochs | 3 | |
|
|
| Learning Rate | 2e-5 | |
|
|
| Batch Size | 16 | |
|
|
| Max Token Length | 256 | |
|
|
| Optimizer | AdamW | |
|
|
| Hardware | 2x NVIDIA Tesla T4 GPUs | |
|
|
| Total FLOPs | 25,605,736,040,851,200 | |
|
|
|
|
|
## Evaluation Metrics |
|
|
|
|
|
| Metric | Value | |
|
|
|------------------|----------| |
|
|
| Accuracy | 0.9656 | |
|
|
| Validation Loss | 0.1513 | |
|
|
| Training Loss | 0.0483 | |
|
|
| Samples/sec | 65.354 | |
|
|
| Training Time | ~1.65 hrs| |
|
|
|
|
|
## Example Inference |
|
|
|
|
|
```python |
|
|
from transformers import pipeline |
|
|
|
|
|
classifier = pipeline("text-classification", model="Elite13/bert-finetuned-mental-health") |
|
|
|
|
|
text = "I'm tired of everything. Nothing makes sense anymore." |
|
|
result = classifier(text) |
|
|
print(result) |
|
|
|
