britannica11/eb1911

Encyclopædia Britannica, 11th Edition (1911) — Full Corpus with Knowledge Graphs

The complete text of the 1911 Encyclopædia Britannica — all ~37,000 articles — as clean Markdown, together with three knowledge graphs reconstructed from the edition and its index volume: a cross-reference graph, a subject taxonomy, and an authorship graph. None of the three exist in the source text; they were rebuilt from the printed classified index and the contributor tables, and they are the reason this is more than a text dump.

Rendered from britannica11.org.

What's in the bundle

file	what it is
articles.jsonl	one record per article — the text as Markdown, plus metadata, sections, and denormalized categories / cross-references / contributors
xref_edges.jsonl	the cross-reference graph — one {from, to, display} edge per resolved internal link
topics.json	the subject taxonomy — the vol-29 classified index as nodes {id, name, path, parent, articles}
contributors.json	the authorship graph — the roster of scholars who wrote the edition, each with the article ids they contributed
manifest.json	version, exact counts, and SHA-256 checksums
schema.json	the JSON Schema for an articles.jsonl record
LICENSE	terms (CC-BY-SA 4.0) and attribution

The article record

{
  "id": "21-0935-poland-POLAND",
  "title": "POLAND",
  "type": "article",
  "volume": 21,
  "page_start": 902,
  "word_count": 42817,
  "url": "https://www.britannica11.org/article/21-0935-poland-POLAND",
  "categories": ["history/europe-(continental)/general"],
  "sections": [{"title": "Polish Literature", "slug": "polish-literature", "level": 1}],
  "contributors": [{"initials": "R. N. B.", "name": "Robert Nisbet Bain"}],
  "images": [{"file": "..."}],
  "xrefs": [{"to": "21-0962-poland-russian-POLAND__RUSSIAN", "display": "Poland, Russian"}],
  "markdown": "(Polish *Polska* …)"
}

Field-by-field types are in schema.json. Images are carried as references (file), not binaries — the corpus stays light for text pipelines, and vision pipelines can fetch the image on demand.

Why the graphs matter

Every existing digitization of the 1911 Britannica is flat OCR text. This one carries structure that had to be reconstructed, not extracted:

• Cross-reference graph (xref_edges.jsonl) — every article joined to the articles it names, resolved and disambiguated (Zürich the city and Zürich the canton are distinct nodes). GraphRAG fuel.
• Subject taxonomy (topics.json) — all ~37,000 articles classified into the edition's own subject hierarchy, taken from the printed vol-29 index. Scope retrieval to a subtree ("everything under Zoology").
• Authorship graph (contributors.json) — the cryptic initials ("R. N. B.") resolved to named scholars with their credentials: a prosopography of 1911 scholarship, and an authority/provenance signal.

Loading

import json

articles = [json.loads(l) for l in open("articles.jsonl", encoding="utf-8")]
edges    = [json.loads(l) for l in open("xref_edges.jsonl", encoding="utf-8")]
topics   = json.load(open("topics.json", encoding="utf-8"))
authors  = json.load(open("contributors.json", encoding="utf-8"))

from datasets import load_dataset
ds = load_dataset("britannica11/eb1911", data_files="articles.jsonl", split="train")

Content note

This is a 1911 encyclopedia. Its science is a century out of date; its geography and politics describe a vanished world; and it contains framings, terms, and views — on race, empire, and much else — that were the educated consensus of its time and are offensive by modern standards. It is preserved and distributed as a historical document, not a current reference. If you train or ground a model on it, treat it accordingly.

Provenance

The text is the 1911 Encyclopædia Britannica (public domain), transcribed by Wikisource and the Internet Archive. This corpus renders that text through a faithful marker-encoded pipeline, and reconstructs the three graphs from the printed vol-29 classified index and the per-volume contributor tables. Full methodology: britannica11.org/about.

License & attribution

Released under CC-BY-SA 4.0. The underlying 1911 text is in the public domain; the transcription is courtesy of Wikisource; the Markdown rendering, the resolved cross-reference and authorship graphs, and the machine-readable subject taxonomy are the contribution of this project. Please attribute as:

Encyclopædia Britannica, 11th Edition corpus — britannica11.org — CC-BY-SA 4.0.

Exact counts, version, and checksums are in manifest.json.