Sathwik3/distilbert-emotion-classifier

DistilBERT Emotion Classifier

Model Description

This model is a fine-tuned version of distilbert-base-uncased for multi-class emotion classification. The model classifies text into different emotional categories, enabling applications in sentiment analysis, customer feedback analysis, and social media monitoring.

Developed by: Sathwik3

Model type: Text Classification (Emotion Detection)

Language(s): English

License: Apache 2.0

Base model: distilbert-base-uncased

Model Details

Architecture

The model is based on DistilBERT, a distilled version of BERT that retains 97% of BERT's language understanding while being 40% smaller and 60% faster. The architecture consists of:

• 6 transformer layers
• 768 hidden dimensions
• 12 attention heads
• ~66M parameters
• Classification head for emotion prediction

Training Objective

The model was fine-tuned using cross-entropy loss for multi-class classification, optimizing for accurate emotion categorization across multiple emotional states.

Intended Uses

Direct Use

The model can be directly used for:

• Emotion detection in text documents
• Sentiment analysis of customer reviews and feedback
• Social media monitoring to understand emotional tone
• Content moderation based on emotional content
• Mental health applications for emotion tracking in journals
• Chatbot enhancement for emotion-aware responses

Downstream Use

This model can be integrated into larger systems for:

• Customer service platforms for automated response routing
• Market research tools for analyzing consumer sentiment
• Educational platforms for emotional intelligence training
• Healthcare applications for mental wellness monitoring

Out-of-Scope Use

The model should not be used for:

• Clinical diagnosis or medical decision-making
• Making critical decisions about individuals without human oversight
• Applications where misclassification could cause harm
• Languages other than English (without additional fine-tuning)
• Real-time crisis intervention or emergency response

Limitations and Bias

Limitations

• Language limitation: The model is trained primarily on English text and may not perform well on other languages or code-switched text
• Context sensitivity: Short texts or texts lacking context may be misclassified
• Domain specificity: Performance may vary across different domains (e.g., formal vs. informal text)
• Sarcasm and irony: The model may struggle with non-literal expressions
• Cultural nuances: Emotion expression varies across cultures, which may affect performance

Bias Considerations

• The model's predictions may reflect biases present in the training data
• Emotion categories may not universally apply across all cultures and contexts
• Performance may vary across demographic groups depending on training data representation
• Users should validate model outputs, especially in sensitive applications

Recommendations

• Always review model predictions in high-stakes applications
• Use the model as a decision support tool, not a sole decision-maker
• Evaluate performance on your specific use case before deployment
• Monitor for bias and fairness issues in production
• Provide clear communication to end users about the model's capabilities and limitations

How to Get Started with the Model

Use the code below to get started with the model:

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

# Load model and tokenizer
model_name = "Sathwik3/distilbert-emotion-classifier"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)

# Example text
text = "I am so happy and excited about this amazing opportunity!"

# Tokenize and predict
inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True, max_length=512)
with torch.no_grad():
    outputs = model(**inputs)
    predictions = torch.nn.functional.softmax(outputs.logits, dim=-1)
    predicted_class = torch.argmax(predictions, dim=-1).item()

print(f"Predicted emotion class: {predicted_class}")
print(f"Confidence scores: {predictions}")

For pipeline usage:

from transformers import pipeline

# Create emotion classification pipeline
emotion_classifier = pipeline("text-classification", model="Sathwik3/distilbert-emotion-classifier")

# Classify emotion
result = emotion_classifier("I am so happy and excited about this amazing opportunity!")
print(result)

Training Details

Training Data

The model was fine-tuned on an emotion classification dataset. Specific dataset details:

• Dataset: Emotion dataset
• Size: 16000
• Emotion categories: ['sadness', 'joy', 'love', 'anger', 'fear', 'surprise']
• Data split: Train,Validation,Test

Training Procedure

Preprocessing

• Text tokenization using DistilBERT tokenizer
• Maximum sequence length: 512 tokens
• Truncation and padding applied as needed

Training Hyperparameters

• Training regime: Mixed precision (fp16)
• Optimizer: AdamW
• Learning rate: 2e-5
• Batch size: 64
• Number of epochs: 2
• Weight decay: 0.01

Evaluation

Testing Data & Metrics

Testing Data

• Test set: [Description of test data - placeholder]
• Test set size: [Number of examples - placeholder]
• Distribution: [Class distribution information - placeholder]

Metrics

The model's performance is evaluated using:

• Accuracy: Overall classification accuracy
• F1 Score: Macro and weighted F1 scores for balanced evaluation
• Precision: Per-class and average precision
• Recall: Per-class and average recall
• Confusion Matrix: For detailed error analysis

Results

Overall Performance

Metric	Value
Accuracy	0.9295
Weighted F1	0.9292

Technical Specifications

Model Architecture

• Base Model: DistilBERT (distilbert-base-uncased)
• Model Size: ~66M parameters (base) + classification head
• Layers: 6 transformer layers
• Hidden Size: 768
• Attention Heads: 12
• Intermediate Size: 3072
• Max Sequence Length: 512 tokens
• Vocabulary Size: 30,522 tokens

Software

• Framework: PyTorch
• Library: Hugging Face Transformers
• Python Version: 3.10
• Key Dependencies:
  • transformers
  • torch
  • tokenizers

Citation

If you use this model in your research or applications, please cite:

BibTeX:

@misc{sathwik3-distilbert-emotion,
  author = {Sathwik3},
  title = {DistilBERT Emotion Classifier},
  year = {2024},
  publisher = {Hugging Face},
  howpublished = {\url{https://huggingface.co/Sathwik3/distilbert-emotion-classifier}}
}

Please also cite the original DistilBERT paper:

@article{sanh2019distilbert,
  title={DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter},
  author={Sanh, Victor and Debut, Lysandre and Chaumond, Julien and Wolf, Thomas},
  journal={arXiv preprint arXiv:1910.01108},
  year={2019}
}

APA:

Sathwik3. (2024). DistilBERT Emotion Classifier. Hugging Face. https://huggingface.co/Sathwik3/distilbert-emotion-classifier

Model Card Authors

Sathwik3

Model Card Contact

For questions or feedback about this model, please open an issue in the model's repository or contact via Hugging Face.

---

This model card follows the guidelines from Mitchell et al. (2019) and the Hugging Face Model Card template.