neeraj-mO/bert-emotion-finetuned

BERT Emotion Classifier (Fine-tuned)

Fine-tuned bert-base-uncased for 6-class emotion classification on the dair-ai/emotion dataset.

Model Description

• Author: Neeraj Gupta (@neeraj)
• Task: Text Classification (Emotion Detection)
• Base model: bert-base-uncased
• Classes: sadness, joy, love, anger, fear, surprise
• Framework: PyTorch + Hugging Face Transformers

Performance

Metric	Score
Accuracy	91.75%
Weighted F1	91.64%

Evaluated on the validation split after training.

Usage

from transformers import pipeline

classifier = pipeline(
    "text-classification",
    model="neeraj-miniOrange/bert-emotion-finetuned"
)

result = classifier("I am so happy today!")
print(result)
# [{'label': 'joy', 'score': 0.98...}]

Training Details

Hyperparameter	Value
Epochs	2
Learning rate	2e-4
Batch size	64
Weight decay	0.01
Optimizer	AdamW

Data split: 70% train / 20% test / 10% validation (stratified)

Limitations

• Trained on short English tweets; may not generalize well to long-form text or other languages.
• Emotion labels are subjective; the model reflects biases present in the training data.
• Not intended for clinical or mental health diagnosis.

Links

• Dataset: dair-ai/emotion
• Base model: bert-base-uncased

Citation

If you use this model, please cite the BERT paper:

@article{devlin2019bert,
  title={BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding},
  author={Devlin, Jacob and Chang, Ming-Wei and Lee, Kenton and Toutanova, Kristina},
  journal={NAACL},
  year={2019}
}