Datasets:

sagorsarker
/

bangla-wikipedia

license: mit
configs:
  - config_name: '20250620'
    data_files:
      - split: train
        path: 20250620/train-*
dataset_info:
  config_name: '20250620'
  features:
    - name: id
      dtype: string
    - name: url
      dtype: string
    - name: title
      dtype: string
    - name: content
      dtype: string
  splits:
    - name: train
      num_bytes: 995443287
      num_examples: 171674
  download_size: 359387209
  dataset_size: 995443287

📚 Bangla Wikipedia Dump

This dataset contains a cleaned and processed version of the Bangla Wikipedia dump, structured for easy use in Natural Language Processing tasks such as language modeling, text classification, and content generation.

Getting Started

To download full datasets:

from datasets import load_dataset

dataset = load_dataset("sagorsarker/bangla-wikipedia")

To download a subset or specific version:

from datasets import load_dataset

dataset = load_dataset("sagorsarker/bangla-wikipedia", data_dir="<subset_name>")
# example
# dataset = load_dataset("sagorsarker/bangla-wikipedia", data_dir="20250620")

📄 Dataset Details

Language: Bangla (Bengali)
Source: Official Wikipedia Bangla dump
License: MIT
Split: train only
Total Examples: 171,674
Total Size: ~995 MB (unzipped)
Wikipedia Version: 20 June 2025

Each sample in the dataset contains:

id: Unique article ID
url: Source Wikipedia URL
title: Article title
content: Cleaned article body text

📦 Configs

configs:
- config_name: '20250620'
  data_files:
  - split: train
    path: 20250620/train-*

💡 Use Cases

This dataset is well-suited for:

Pretraining/fine-tuning language models for Bangla
Information retrieval or summarization tasks
Next-word or masked-word prediction models
Bangla text classification, QA, or conversational agents

🤝 Contribution

This is a philanthropic open-source contribution to support the growth of the Bangla NLP community. Contributions and feedback are warmly welcome!