sayurio/bangla-site-scrape

Bangladeshi Web Scraped Dataset

---

Dataset Description

This dataset is a comprehensive collection of scraped web data from various Bangladeshi websites. It is designed to facilitate natural language processing (NLP) tasks for the Bengali language, including analysis of e-commerce trends, news classification, and general text generation.

The data is provided in JSONL (JSON Lines) format, making it easy to parse and integrate into machine learning pipelines.

Key Features

• Diverse Sources: Includes data from e-commerce platforms, news portals, blogs, and informational articles.
• Rich Metadata: E-commerce entries include specific fields such as pricing, product descriptions, and specifications.
• Open Access: Free to download and use under the MIT License.

---

Dataset Structure

Each line in the JSONL file represents a single scraped page or item. The structure varies slightly depending on the source type (e.g., e-commerce vs. news article).

General Structure

{
  "url": "[https://example.com/article](https://example.com/article)",
  "domain": "example.com",
  "scrape_date": "2023-10-27",
  "category": "news",
  "content": "Full text content of the article...",
  "title": "Article Headline"
}

Gemini said Here is a README.md template tailored for your Hugging Face dataset.

README.md Markdown

license: mit task_categories: - text-generation - text-classification language: - bn size_categories: - 10K<n<100K pretty_name: Bangladeshi Web Scraped Dataset tags: - e-commerce - news - articles - bangladesh - web-scraping

Bangladeshi Web Scraped Dataset

Table of Contents

• Dataset Description
• Dataset Structure
• Data Fields
• Usage
• Licensing
• Citation

---

Dataset Description

This dataset is a comprehensive collection of scraped web data from various Bangladeshi websites. It is designed to facilitate natural language processing (NLP) tasks for the Bengali language, including analysis of e-commerce trends, news classification, and general text generation.

The data is provided in JSONL (JSON Lines) format, making it easy to parse and integrate into machine learning pipelines.

Key Features

• Diverse Sources: Includes data from e-commerce platforms, news portals, blogs, and informational articles.
• Rich Metadata: E-commerce entries include specific fields such as pricing, product descriptions, and specifications.
• Open Access: Free to download and use under the MIT License.

---

Dataset Structure

Each line in the JSONL file represents a single scraped page or item. The structure varies slightly depending on the source type (e.g., e-commerce vs. news article).

General Structure

{
  "url": "[https://example.com/article](https://example.com/article)",
  "domain": "example.com",
  "scrape_date": "2023-10-27",
  "category": "news",
  "content": "Full text content of the article...",
  "title": "Article Headline"
}

Data Fields

Field	Type	Description
url	string	The source URL of the scraped data.
domain	string	The website domain (e.g., prothomalo.com, daraz.com.bd).
category	string	Type of content (e.g., news, e-commerce, article).
title / product_name	string	The headline of the article or name of the product.
content / description	string	The main body text or product description.
price	string	(E-commerce only) The price of the item, usually including currency.
scrape_date	string	The date the data was collected.

You can load this dataset directly using the Hugging Face datasets library:

Python

from datasets import load_dataset

dataset = load_dataset("your-username/your-dataset-name")

# Inspect the first example
print(dataset['train'][0])

Potential Use Cases

-Bengali NLP: Training language models on contemporary Bangladeshi web text.

• Sentiment Analysis: Analyzing news or product reviews.
• Market Analysis: Tracking e-commerce pricing and product trends in Bangladesh.
• Classification: Categorizing web content by domain or topic.

Licensing

This dataset is released under the MIT License.

Disclaimer: The data contained in this dataset is scraped from publicly available websites. Users are responsible for ensuring that their use of this data complies with the terms of service of the source websites and applicable local laws.