EComReviews-BD

Each review was manually annotated with an overall sentiment label (positive, neutral, or negative) based on the expressed opinion. Additionally, category-specific aspect-level sentiment annotations were assigned following predefined annotation guidelines to support aspect-based sentiment analysis (ABSA). To facilitate multilingual research, an English-translated version of the dataset was generated using the NLLB-200 (facebook/nllb-200-distilled-600M) neural machine translation model while preserving the original sentiment annotations.

The final dataset is provided as Microsoft Excel files containing marketplace information, product category, product name, brand, price, overall rating, review text, overall sentiment labels, and aspect-level sentiment annotations, making it suitable for reproducible research in multilingual sentiment analysis, ABSA, natural language processing, machine learning, and large language models.