MindMate Text Emotion Model
Fine-tuned XLM-RoBERTa-base for multilingual emotion detection.
Trained on
- DAIR-AI Emotion dataset
- Additional small correction dataset (Malayalam + others)
Performance (test set)
- Overall Accuracy: 92.70%
- Detailed metrics:
- sadness: precision 0.97, recall 0.96, f1 0.97
- joy: 0.94 / 0.97 / 0.95
- anger: 0.92 / 0.92 / 0.92
- love: 0.92 / 0.74 / 0.82
- fear: 0.88 / 0.89 / 0.88
- surprise: 0.68 / 0.80 / 0.74
Supported languages
- Excellent English performance
- Strong Malayalam support (tested in notebook)
Usage example
from transformers import pipeline
classifier = pipeline("text-classification", model="sidharths9105/MindMate-Text-Emotion-Model")
print(classifier("ഞാൻ വളരെ സന്തോഷവാനാണ്")) # → joy
print(classifier("I feel very angry today")) # → anger