arlette80-laure/hotel-sentiment-distilbert

Hotel Sentiment DistilBERT

This model is a fine-tuned version of DistilBERT created to classify hotel customer reviews as POSITIVE or NEGATIVE.

Model Description

The model was developed as a practical Hugging Face fine-tuning project.

It demonstrates how a pretrained Transformer model can be adapted to a specific text-classification task using a custom dataset.

Base Model

The base model is:

distilbert/distilbert-base-uncased

DistilBERT is a smaller and faster version of BERT.

Task

The model performs binary sentiment classification.

The two classes are:

• NEGATIVE
• POSITIVE

Training Dataset

For this educational project, the model was fine-tuned using a small custom dataset of hotel customer reviews.

Example positive review:

The room was clean and comfortable.

Label:

POSITIVE

Example negative review:

The room was dirty and the service was terrible.

Label:

NEGATIVE

Intended Use

The model can be used to demonstrate automatic sentiment analysis of hotel and hospitality customer reviews.

Example applications include:

• Hotel review analysis
• Customer feedback classification
• Hospitality sentiment monitoring
• AI and NLP training demonstrations

How to Use

After the model files have been uploaded to this repository, the model can be loaded with Hugging Face Transformers:

from transformers import pipeline

classifier = pipeline(
    "text-classification",
    model="arlette80-laure/hotel-sentiment-distilbert"
)

result = classifier(
    "The hotel was excellent and the staff were very friendly."
)

print(result)

Labels

ID	Label
0	NEGATIVE
1	POSITIVE

Limitations

This model was created for educational purposes using a very small training dataset.

Therefore:

• It should not be considered production-ready.
• It may perform poorly on complex or ambiguous reviews.
• It may have difficulty with sarcasm.
• It may have difficulty with mixed positive and negative sentiment.
• It was trained primarily using English examples.
• More training data would be required for reliable real-world deployment.

Future Improvements

Future versions can be improved by:

• Increasing the number of training examples
• Using real hotel customer reviews
• Creating separate training, validation, and test datasets
• Measuring precision, recall and F1-score
• Supporting French and Spanish reviews
• Performing hyperparameter tuning
• Testing the model on unseen real-world data

Framework

This model was developed using:

• Hugging Face Transformers
• Hugging Face Datasets
• PyTorch
• Hugging Face Trainer

License

MIT

Purpose

This repository forms part of a hands-on Hugging Face learning project demonstrating the complete workflow:

Pretrained Model → Custom Dataset → Tokenization → Fine-tuning → Evaluation → Model Publishing → Inference