README.md

metadata

language:
  - bn
dataset_info:
  features:
    - name: text
      dtype: string
    - name: label
      dtype:
        class_label:
          names:
            '0': Positive
            '1': Neutral
            '2': Negative
  splits:
    - name: train
      num_bytes: 72611.93116395494
      num_examples: 559
    - name: validation
      num_bytes: 15587.534418022527
      num_examples: 120
    - name: test
      num_bytes: 15587.534418022527
      num_examples: 120
  download_size: 46964
  dataset_size: 103786.99999999999
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*
      - split: test
        path: data/test-*

Bengali Sentiment Analysis Dataset

---

This dataset contains YouTube comments in Bangla, labeled with sentiment categories. It has been meticulously annotated and split into training, validation, and test sets. The dataset is ideal for various natural language processing (NLP) tasks such as sentiment analysis, text classification, and language understanding.

Dataset Structure

The dataset is organized as follows:

• Train Set: 559 comments
• Validation Set: 120 comments
• Test Set: 120 comments

Each entry in the dataset has the following features:

1. Comment: The Bangla text comment extracted from YouTube videos.
2. Sentiment: The sentiment label associated with each comment. It can be one of the following three categories:
   • positive
   • negative
   • neutral

Dataset Details

• Total Rows: 799 comments
• Features:
  • Comment (string): The text of the comment.
  • Sentiment (string): The sentiment label indicating whether the comment is positive, negative, or neutral.

The data distribution ensures a balanced representation of each sentiment class, making it suitable for training and evaluation of sentiment analysis models.

Annotations

The sentiment labels were annotated manually to ensure the highest possible quality. The annotation guidelines considered the tone, context, and overall sentiment of each comment to assign one of the three sentiment classes.

Splits Information

The dataset is split into three sets:

Split	Number of Samples
Train	559
Validation	120
Test	120

This split was created to provide a sufficient amount of data for training and evaluation while ensuring robust performance evaluation on unseen samples.

Citation

@article{Mirza2024, title={BengaliSent: Bangla Sentiment Analysis Dataset}, author={Mirza Abbas Uddin}, year={2024} }

Usage Example

The dataset can be loaded using the Hugging Face datasets library as shown below:

from datasets import load_dataset

# Load the dataset from the Hugging Face Hub
dataset = load_dataset("mirzaaa10/bengali_sent")

# Access the training data
train_data = dataset['train']

# Print a sample
print(train_data[0])