A Bangladeshi Movie Review Dataset in Bangla, English, and Banglish Language with Sentiment and Genre Labels

This dataset contains 12,000 unique movie reviews carefully collected and designed to support research in Sentiment Analysis, Natural Language Processing (NLP), and Artificial Intelligence. The reviews are written in Bangla, English, and Banglish, reflecting the natural way Bangladeshi audiences express their opinions about movies on online platforms.

To make the dataset realistic and representative of real-world user-generated content, the reviews vary in length, tone, and writing style. Some reviews are short, single-line opinions, while others are longer and include mixed expressions, emotions, and personal viewpoints, similar to comments commonly found on YouTube and social media platforms.

Each review in the dataset has been manually annotated with the following labels:
• Sentiment labels: Positive, Negative, or Neutral
• Genre labels: Comedy–Drama, Liberation War, Action, Thriller, and Romantic

Special care was taken to maintain logical and semantic consistency throughout the annotation process. For example, reviews expressing clear appreciation, praise, or excitement were not labeled as Negative, and reviews showing dissatisfaction or criticism were not marked as Positive. Genre labels were assigned based on the officially recognized category of each movie to ensure accuracy and clarity.

All reviews were manually collected, verified, and cleaned, and duplicate or irrelevant entries were removed to improve data quality. No personally identifiable information (PII), such as usernames or profile links, is included in the dataset, ensuring ethical use of publicly available data.

This dataset provides a reliable and well-structured resource for training, testing, and evaluating NLP and machine learning models, particularly for sentiment classification, genre-wise opinion analysis, and AI-based movie recommendation systems in a low-resource language context