EncycNet-Wikipedia Corpus: A Trilingual (EN/DE/FR) Wikipedia Subset

This dataset provides the full text of Wikipedia articles in three languages (German, English, French) for all entities linked to entries in the EncycNet corpus (https://encycnet.github.io/) — a knowledge graph built from 19th- and early 20th-century German encyclopedias. Each entry corresponds to a single concept/entity and includes the article text in whichever of the three languages a Wikipedia page exists, along with the IDs of the corresponding EncycNet entries it is linked to. As the underlying EncycNet corpus is drawn from encyclopedias published between the early 19th and early 20th century, the linked entities are naturally historical in nature, which means this dataset does not cover living people, contemporary media, or other present-day topics. The articles were retrieved from Wikipedia in January 2025. Coverage: 85,433 German Wikipedia pages 65,836 English Wikipedia pages 59,271 French Wikipedia pages Data format:The dataset is distributed as a single JSON file: a dictionary keyed by article identifier (prefixed "article_"), where each value has the structure: { "article_<lemma>": { "languages": { "de": "<German text>", "en": "<English text>", "fr": "<French text>" }, "corpus_ids": ["<id1>", "<id2>", "..."] }, ...} The key (e.g. "article_Aal) is the Wikipedia article lemma (German), prefixed with "article_". languages: full article text (including any residual wikitext section headers, e.g. "== Weblinks ==") per language; the value is an empty string if no Wikipedia page exists in that language for the entity. corpus_ids: identifiers of the corresponding entries in the EncycNet knowledge graph (https://encycnet.github.io/), of the form "<node-id><Encyclopedia>-<Year>" (e.g. "d2e42Brockhaus-1911"), linking each Wikipedia article to the specific historical encyclopedia entry/entries it was aligned with. A single article may map to multiple corpus_ids where several historical encyclopedias covered the same concept. Wikipedia article text is © Wikipedia contributors and is redistributed under CC BY-SA 4.0 (and GFDL), in line with Wikipedia's licensing terms.