aurora-bert-sentiment
Sentiment analysis model based on BERT, fine-tuned on product reviews.
Model Details
Trained on a proprietary dataset of 500k product reviews. Input is a text snippet, output is a sentiment label (positive/neutral/negative).
Intended Use
Designed for analyzing customer feedback in the Aurora retail platform. Not intended for medical, legal, or financial decisions.
Limitations and Bias
May exhibit bias toward informal language. Performance degrades on non-English text. Users should validate on their own data.
Training Data
Trained on Aurora internal review data collected between 2023 and 2025.