Multilabeled Bengali Sentiment and Emotion Classification Dataset

This dataset, titled Multilabeled Sentiment and Emotion Classification Dataset, is developed to advance natural language processing (NLP) research in the Bengali language, particularly in the domains of sentiment analysis and emotion detection. It contains 40,811 entries of user-generated text from Bengali social media and comment sections, each annotated with two labels: one for sentiment and another for emotion.

Sentiment Categories:
The dataset is annotated with five sentiment classes:

Very Negative – 8,979 entries (21.9%)

Negative – 10,757 entries (26.3%)

Neutral – 8,662 entries (21.2%)

Positive – 7,186 entries (17.8%)

Very Positive – 5,227 entries (12.8%)

This distribution highlights a larger presence of negative and neutral sentiments, indicating a critical tone in the source data.

Emotion Categories:
There are seven emotion classes in the dataset:

Happy – 8,154 entries (19.9%)

Surprised – 3,470 entries (8.5%)

Sexual – 7,250 entries (17.7%)

Religious – 2,449 entries (6.0%)

Calm – 7,583 entries (18.5%)

Hateful – 4,919 entries (12.0%)

Fearful – 7,086 entries (17.3%)

This emotion distribution reveals that Happy, Calm, and Sexual emotions are among the most prevalent, while Religious and Surprised emotions are relatively less represented.

Applications:
This dataset is suitable for training and evaluating machine learning and deep learning models for tasks such as:

Multilabel text classification

Emotion recognition

Sentiment analysis

Hate speech and toxic comment detection in the Bengali language

Language:
Bengali (Bangla)