c-ram/roberta-amazon-reviews-sentiment

RoBERTa Amazon Reviews Sentiment Classifier

Fine-tuned version of roberta-base for 3-class sentiment classification (negative / neutral / positive) on Amazon product reviews.

Model Details

• Base model: roberta-base
• Task: Sentiment classification (3-class)
• Dataset: Amazon product reviews (balanced sample)
• Language: English
• Repository: amazon-review-classifier-roberta-sentiment-finetuning

Performance

Evaluated on a held-out test set of 1,086 reviews:

Class	Precision	Recall	F1-Score	Support
Negative	0.84	0.85	0.85	361
Neutral	0.77	0.71	0.74	350
Positive	0.86	0.91	0.89	375
Macro avg	0.82	0.82	0.82	1086
Weighted avg	0.82	0.83	0.83	1086

Overall accuracy: 83%

Usage

from transformers import pipeline

classifier = pipeline("text-classification", model="c-ram/roberta-amazon-reviews-sentiment")
classifier("This product exceeded my expectations!")