empathy

Therapy Chatbot - Mental Health Intent Classification

This model classifies mental health-related text into 11 different intent categories.

Model Description

• Model type: DistilBERT for Sequence Classification
• Language: English
• Training data: 3,508 therapy context-response pairs
• Task: Intent classification for mental health chatbot

Intent Categories

The model classifies text into these 11 intents:

• depression - Feelings of sadness, hopelessness, worthlessness
• anxiety - Worry, panic, nervousness
• self_esteem - Issues with self-worth and confidence
• relationship - Relationship problems and conflicts
• family - Family-related concerns
• sleep_issues - Insomnia and sleep problems
• anger - Anger management issues
• suicide - Suicidal thoughts (crisis detection)
• trauma - PTSD and trauma-related issues
• grief - Loss and bereavement
• general_support - General emotional support needs

Usage

from transformers import DistilBertTokenizerFast, DistilBertForSequenceClassification
import torch

# Load model and tokenizer
model = DistilBertForSequenceClassification.from_pretrained("YOUR-USERNAME/therapy-chatbot")
tokenizer = DistilBertTokenizerFast.from_pretrained("YOUR-USERNAME/therapy-chatbot")

# Make prediction
text = "I feel worthless and can't sleep at night"
inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True)

with torch.no_grad():
    outputs = model(**inputs)

prediction = torch.argmax(outputs.logits, dim=1).item()
print(f"Predicted intent: {prediction}")

Training

• Epochs: 6
• Batch size: 16
• Learning rate: 3e-5
• Validation loss: ~0.030

Limitations

⚠️ Important: This model is NOT a replacement for professional mental health care. It's designed to assist in categorizing user messages for appropriate responses.

Disclaimer

This is an AI model for educational and support purposes only. In case of mental health emergencies, please contact:

• National Suicide Prevention Lifeline: 988
• Crisis Text Line: Text HOME to 741741