TheOneWhoWill/naruto-boruto-wiki

Naruto & Boruto Wiki Dataset

A structured text dataset containing all articles from the Naruto and Boruto Fandom wikis. Extracted via the MediaWiki API with clean, readable markdown-formatted text suitable for LLM pretraining, fine-tuning, or knowledge-augmented tasks.

Dataset Summary

This dataset contains 8,705 articles (7,934 from Naruto Wiki + 771 from Boruto Wiki) totaling 23.7 million characters (5.9M estimated tokens). Each article is stored as clean markdown-style text with metadata including authorship, timestamps, and page IDs.

The Naruto and Boruto wikis are comprehensive fan-maintained encyclopedias covering characters, episodes, jutsu, locations, lore, and more from the Naruto and Boruto anime/manga franchises. This dataset captures the full textual content of both wikis as of June 2026.

Key Statistics

Metric	Value
Total articles	8,705
Naruto Wiki articles	7,934
Boruto Wiki articles	771
Total characters	23,653,963
Avg. characters/article	2,717
Est. tokens (~4 char/tok)	~5.9M
Compressed JSONL size	~26 MB

Data Format

The dataset is provided as a single JSONL file (naruto_boruto_wiki.jsonl). One JSON object per line, with the following schema:

{
  "title": "Naruto Uzumaki",
  "wiki": "naruto",
  "url": "https://naruto.fandom.com/wiki/Naruto_Uzumaki",
  "text": "# Naruto Uzumaki\n\nNaruto Uzumaki (うずまきナルト, Uzumaki Naruto) is a shinobi of Konohagakure...",
  "metadata": {
    "page_id": 1526,
    "created_at": "2006-10-09T00:51:16Z",
    "created_by": "64.222.16.224",
    "last_modified": "2026-06-06T02:37:08Z",
    "last_modified_by": "SomeUser",
    "length_bytes": 256907
  }
}

Fields

Field	Type	Description
title	string	Article title (page name)
wiki	string	Source wiki: "naruto" or "boruto"
url	string	Full URL to the wiki article
text	string	Clean markdown-formatted article text
metadata.page_id	int	MediaWiki page ID
metadata.created_at	string	ISO 8601 timestamp of first revision
metadata.created_by	string	Username/IP of page creator
metadata.last_modified	string	ISO 8601 timestamp of most recent revision
metadata.last_modified_by	string	Username of last editor
metadata.length_bytes	int	Raw wikitext length in bytes

Text Formatting

The text field contains markdown-style formatting:

• # Title through ###### Subheading for headings
• - List item for bullet lists
• Paragraphs separated by blank lines
• Bold/italic text is preserved as plain text (no markdown emphasis markers)
• Tables, infoboxes, figures, navigation elements, and reference superscripts are removed
• External links are stripped to their display text only

Intended Uses

• LLM Pretraining: Domain-specific knowledge injection for anime/fiction-aware models
• Fine-tuning: Instruction tuning or continued pretraining on structured wiki knowledge
• Knowledge Retrieval: Testing RAG systems on fictional universe knowledge
• Text Generation: Training models to generate wiki-style articles about fictional topics
• Tokenization: Training custom tokenizers on domain-specific vocabulary (jutsu names, character names, etc.)

Data Sources

• Naruto Wiki: naruto.fandom.com — ~8,019 pages indexed
• Boruto Wiki: boruto.fandom.com — ~1,500+ pages indexed

Data was collected using the public MediaWiki API:

• action=query&list=allpages — page listing
• action=parse — HTML content retrieval
• action=query&prop=revisions|info — metadata retrieval

Limitations & Biases

• Incomplete coverage: A small number of pages may have been skipped due to API errors or access restrictions
• Fandom bias: Content reflects the collective knowledge and biases of wiki contributors; some articles may contain inaccuracies, speculation, or outdated information
• English only: Only English-language articles are included
• Temporal snapshot: Content reflects the state of both wikis as of June 2026; newer episodes, chapters, or edits are not included
• No images/media: All visual content (images, videos, galleries) is excluded; only textual content is preserved
• Structured data loss: Infoboxes, tables, and other structured wiki elements are excluded, which may omit factual information (e.g., character stats, episode lists)

License

Content is sourced from Fandom wikis, which are licensed under CC BY-SA 3.0. Any use of this dataset must comply with the same license terms.

Creation Process

1. Page listing — Retrieved all page titles via allpages API with pagination
2. HTML extraction — Fetched parsed HTML for each page via parse API
3. Metadata collection — Retrieved first/last revision info via revisions API
4. Text conversion — Parsed HTML with Python's html.parser, extracting paragraphs, headings, and list items while stripping tables, figures, and navigation elements
5. Deduplication — Progress-tracking ensured no duplicate articles
6. Output — Articles saved to a single JSONL file

The scraping script (scrape_wikis.py) is included in the dataset repository and can be used to reproduce or update the dataset.

Citation

If you use this dataset, please cite:

@dataset{naruto_boruto_wiki_2026,
  author       = {Naruto Wiki Contributors and Boruto Wiki Contributors},
  title        = {Naruto \& Boruto Wiki Dataset},
  year         = {2026},
  publisher    = {Fandom},
  note         = {Compiled via MediaWiki API from naruto.fandom.com and boruto.fandom.com},
  url          = {https://naruto.fandom.com/}
}