Arjo12/distilbert-imdb-sentiment

DistilBERT IMDb Sentiment Analysis

This model is a fine-tuned version of distilbert/distilbert-base-uncased for binary sentiment classification of English movie reviews.

It predicts whether a movie review expresses a positive or negative sentiment.

Model Details

• Model: DistilBERT Base Uncased
• Task: Sentiment Analysis
• Language: English
• Dataset: IMDb Movie Reviews
• Number of classes: 2
• Labels:
  • 0: NEGATIVE
  • 1: POSITIVE
• Maximum sequence length: 256 tokens
• Fine-tuning framework: Hugging Face Transformers
• Training approach: Supervised fine-tuning

Intended Use

This model is intended for:

• Classifying movie reviews as positive or negative.
• Experimenting with transformer-based text classification.
• Learning and demonstrating the Hugging Face fine-tuning workflow.

It is not specifically designed for sarcasm detection, emotion classification, or sentiment analysis in domains other than movie reviews.

Training Data

The model was fine-tuned using the IMDb dataset, which contains movie reviews labeled as positive or negative.

The original training split was divided into:

• Training set: 90% of the original training data.
• Validation set: 10% of the original training data.
• Test set: The original IMDb test split.

The validation set was used for model evaluation and model selection during training.

Training Procedure

The model was fine-tuned using the Hugging Face Trainer API.

Hyperparameter	Value
Base model	distilbert/distilbert-base-uncased
Learning rate	2e-5
Training epochs	3
Training batch size	16
Weight decay	0.01
Maximum sequence length	256
Evaluation strategy	Per epoch
Model selection metric	F1-score
Optimizer	Trainer default

Evaluation

The model was evaluated on the IMDb test split using the following metrics:

• Accuracy
• Precision
• Recall
• F1-score

Metric	Score
Accuracy	0.91176
Precision	0.9099235425294679
Recall	0.914
F1-score	0.9119572158365262

Usage

You can load the model directly using the Transformers pipeline.

from transformers import pipeline

model_id = "YOUR_USERNAME/distilbert-imdb-sentiment"

classifier = pipeline(
    "text-classification",
    model=model_id
)

text = "The movie was fantastic. The acting and story were amazing!"

result = classifier(text)

print(result)

Example output:

[{'label': 'POSITIVE', 'score': 0.99}]

The score is the model's confidence for its predicted label. This example is illustrative, not a measured result.

Limitations and Bias

• The model is trained on IMDb movie reviews and may not generalize well to other types of text.
• It may struggle with sarcasm, ambiguous opinions, mixed sentiment, or complex language.
• Its predictions may reflect biases present in the training data.
• The confidence score should not be interpreted as a perfectly calibrated probability.

Frameworks

• Transformers
• Datasets
• PyTorch

Author

Arjo Ghosh

This project was created as a practical exercise in fine-tuning and deploying transformer-based models using Hugging Face.