aqsaahsan/sentiment-distilbert

Movie Review Sentiment (DistilBERT)

DistilBERT fine-tuned on Rotten Tomatoes movie reviews to classify text as POSITIVE or NEGATIVE.

Model details

• Base model: distilbert/distilbert-base-uncased
• Task: binary sentiment classification
• Language: English

Training data

cornell-movie-review-data/rotten_tomatoes: 10,662 reviews (8,530 train / 1,066 validation / 1,066 test), balanced classes.

Training

• Epochs: 3
• Learning rate: 2e-5
• Batch size: 16
• Max sequence length: 128
• Hardware: Google Colab T4 GPU

Results

Epoch	Training loss	Validation loss	Validation accuracy
1	0.423	0.387	83.1%
2	0.255	0.376	85.5%
3	0.160	0.495	85.4%

Test accuracy: 83.6%

How to use

from transformers import pipeline

clf = pipeline("text-classification", model="aqsaahsan/sentiment-distilbert")
print(clf("This movie was fantastic!"))

Limitations

• The model only has two classes, so neutral text is forced into positive or negative.
• It was trained on short movie reviews, so it may be less accurate on other topics.
• It may carry biases inherited from DistilBERT's pretraining data.