README.md

metadata

language:
  - en
license: cc-by-sa-3.0
tags:
  - treecorpus
  - wikipedia
  - encyclopedia
  - knowledge-base
  - factual-knowledge
  - training-data
  - conversational-ai
  - nlp
  - language-model
  - text-corpus
  - qa-dataset
  - structured-data
  - large-scale
pretty_name: 'TreeCorpus: Wikipedia Knowledge for AI Models'
size_categories:
  - 10M<n<100M

TreeCorpus

TreeCorpus is a comprehensive, structured dataset derived from the latest Wikipedia dumps, specially processed to serve as high-quality training data for conversational AI models. This dataset transforms Wikipedia's encyclopedic knowledge into a format optimized for natural language understanding and generation tasks.

Dataset Statistics

• Size: 26.27 GB (26,272,580,250 bytes)
• Examples: 2,882,766 articles
• Download Size: 13.33 GB (13,326,529,312 bytes)
• Language: English

Data Structure

Each entry in the dataset contains:

• id (string): Unique Wikipedia article identifier
• title (string): Article title
• text (string): Clean, processed text content
• url (string): Source Wikipedia URL
• timestamp (string): Processing timestamp

Key Features

• Clean, Structured Content: Meticulously processed to remove markup, templates, references, and other non-content elements while preserving the informational value of Wikipedia articles.
• Rich Metadata: Each entry includes article ID, title, clean text content, source URL, and timestamp.
• Comprehensive Coverage: Incorporates the full spectrum of Wikipedia's knowledge base, spanning nearly 3 million articles across countless topics.
• Conversational Optimization: Content is processed specifically to support training of dialogue systems, conversational agents, and knowledge-grounded language models.
• Regular Updates: Built from the latest Wikipedia dumps to ensure current information.

Usage

This dataset is ideal for:

• Training large language models requiring broad knowledge bases
• Fine-tuning conversational agents for knowledge-intensive tasks
• Question-answering systems that need factual grounding
• Research in knowledge representation and retrieval in natural language

License and Citation

TreeCorpus is derived from Wikipedia content available under the CC BY-SA 3.0 license. When using this dataset, please provide appropriate attribution to both this dataset and Wikipedia.

Dataset Configuration

The dataset is configured with a default split:

• Split name: train
• Data files pattern: data/train-*

Creation Process

TreeCorpus was created using a specialized pipeline that:

1. Downloads the latest Wikipedia dumps
2. Processes XML content to extract articles
3. Cleans and standardizes text by removing markup, templates, and non-content elements
4. Structures data in a consistent, machine-readable format
5. Filters out redirects, stubs, and non-article content

For more details on the methodology and processing pipeline, please see the accompanying code documentation.