|
--- |
|
library_name: transformers |
|
tags: |
|
- conversation |
|
- empathetic |
|
- roberta-base |
|
license: apache-2.0 |
|
datasets: |
|
- facebook/empathetic_dialogues |
|
language: |
|
- en |
|
metrics: |
|
- accuracy |
|
- f1 |
|
- precision |
|
- recall |
|
base_model: |
|
- FacebookAI/roberta-base |
|
pipeline_tag: text-classification |
|
--- |
|
# Model Card: RoBERTa Fine-Tuned on Empathetic Dialogues |
|
|
|
## Model Description |
|
|
|
This is a RoBERTa-based model fine-tuned on the Empathetic Dialogues dataset for conversational emotion classification. The model leverages the powerful RoBERTa architecture to understand and classify emotional contexts in conversational text. |
|
|
|
### Emotion Classes |
|
The model is trained to classify conversations into the following emotional categories: |
|
- Surprised |
|
- Angry |
|
- Sad |
|
- Joyful |
|
- Anxious |
|
- Hopeful |
|
- Confident |
|
- Disappointed |
|
|
|
### Model Details |
|
- **Base Model**: roberta-base |
|
- **Task**: Emotion Classification in Conversations |
|
- **Dataset**: Empathetic Dialogues |
|
- **Training Approach**: Full Fine-Tuning |
|
- **Number of Emotion Classes**: 8 |
|
|
|
### Model Performance |
|
|
|
| Metric | Score | |
|
|--------|-------| |
|
| Test Loss | 0.8107 | |
|
| Test Accuracy | 73.01% | |
|
| Test F1 Score | 72.96% | |
|
| Runtime | 10.99 seconds | |
|
| Samples per Second | 61.68 | |
|
| Steps per Second | 1.001 | |
|
|
|
## Usage |
|
|
|
### Hugging Face Transformers Pipeline |
|
|
|
```python |
|
from transformers import pipeline |
|
|
|
# Initialize the emotion classification pipeline |
|
classifier = pipeline( |
|
"text-classification", |
|
model="Sidharthan/roberta-base-conv-emotion" |
|
) |
|
|
|
# Classify emotion in a conversation |
|
text = "I'm feeling really frustrated with work lately." |
|
result = classifier(text) |
|
print(result) |
|
``` |
|
|
|
### Direct Model Loading |
|
|
|
```python |
|
from transformers import AutoModelForSequenceClassification, AutoTokenizer |
|
import torch |
|
|
|
# Load the model and tokenizer |
|
model_name = "Sidharthan/roberta-base-conv-emotion" |
|
model = AutoModelForSequenceClassification.from_pretrained(model_name) |
|
tokenizer = AutoTokenizer.from_pretrained(model_name) |
|
|
|
# Prepare input |
|
text = "I'm feeling really frustrated with work lately." |
|
inputs = tokenizer(text, return_tensors="pt") |
|
|
|
# Predict |
|
with torch.no_grad(): |
|
outputs = model(**inputs) |
|
predictions = torch.softmax(outputs.logits, dim=1) |
|
predicted_class = torch.argmax(predictions, dim=1) |
|
``` |
|
|
|
## Limitations |
|
- Performance may vary with out-of-domain conversational contexts |
|
- Emotion classification limited to the 8 specified emotional categories |
|
- Relies on the specific emotional nuances in the Empathetic Dialogues dataset |
|
- Requires careful interpretation in real-world applications |
|
|
|
## Ethical Considerations |
|
- Emotion classification can be subjective |
|
- Potential for bias based on training data |
|
- Should not be used for making critical decisions about individuals |
|
|
|
## License |
|
Apache 2.0 |
|
|
|
## Citations |
|
```bibtex |
|
@misc{roberta-base-conv-emotion, |
|
title={RoBERTa Fine-Tuned on Empathetic Dialogues}, |
|
author={Sidharthan}, |
|
year={2024}, |
|
publisher={Hugging Face} |
|
} |
|
``` |
|
|
|
## Contact |
|
For more information, please contact the model's author. |
