Datasets:

Meddies
/

wiki

license: cc-by-sa-4.0
language:
  - de
  - en
  - es
  - fil
  - fr
  - id
  - ja
  - ko
  - lo
  - ms
  - my
  - pt
  - ru
  - ta
  - th
  - vi
  - zh
task_categories:
  - text-retrieval
  - feature-extraction
pretty_name: Meddies Multilingual Wikipedia Cirrus 20260628
configs:
  - config_name: burmese
    data_files:
      - split: train
        path: burmese/*.parquet
  - config_name: chinese
    data_files:
      - split: train
        path: chinese/*.parquet
  - config_name: english
    data_files:
      - split: train
        path: english/*.parquet
  - config_name: filipino
    data_files:
      - split: train
        path: filipino/*.parquet
  - config_name: french
    data_files:
      - split: train
        path: french/*.parquet
  - config_name: german
    data_files:
      - split: train
        path: german/*.parquet
  - config_name: indonesian
    data_files:
      - split: train
        path: indonesian/*.parquet
  - config_name: japanese
    data_files:
      - split: train
        path: japanese/*.parquet
  - config_name: korean
    data_files:
      - split: train
        path: korean/*.parquet
  - config_name: laos
    data_files:
      - split: train
        path: laos/*.parquet
  - config_name: malay
    data_files:
      - split: train
        path: malay/*.parquet
  - config_name: portuguese
    data_files:
      - split: train
        path: portuguese/*.parquet
  - config_name: russian
    data_files:
      - split: train
        path: russian/*.parquet
  - config_name: spanish
    data_files:
      - split: train
        path: spanish/*.parquet
  - config_name: tamil
    data_files:
      - split: train
        path: tamil/*.parquet
  - config_name: thai
    data_files:
      - split: train
        path: thai/*.parquet
  - config_name: vietnamese
    data_files:
      - split: train
        path: vietnamese/*.parquet

Meddies Multilingual Wikipedia

Parquet export of Wikipedia namespace-0 articles from the official Wikimedia CirrusSearch content index dump 20260628.

Rows: 14,789,988.

Subsets

Language	Config	Code	Source index	Rows	Shards
Burmese	`burmese`	`my`	`mywiki_content`	84,709	1
Chinese	`chinese`	`zh`	`zhwiki_content`	862,880	8
English	`english`	`en`	`enwiki_content`	3,255,704	64
Filipino / Tagalog	`filipino`	`fil`	`tlwiki_content`	28,230	1
French	`french`	`fr`	`frwiki_content`	1,835,704	20
German	`german`	`de`	`dewiki_content`	2,072,166	20
Indonesian	`indonesian`	`id`	`idwiki_content`	419,640	3
Japanese	`japanese`	`ja`	`jawiki_content`	1,042,541	14
Korean	`korean`	`ko`	`kowiki_content`	349,707	3
Lao	`laos`	`lo`	`lowiki_content`	3,959	1
Malay	`malay`	`ms`	`mswiki_content`	387,597	1
Portuguese	`portuguese`	`pt`	`ptwiki_content`	801,255	7
Russian	`russian`	`ru`	`ruwiki_content`	998,978	22
Spanish	`spanish`	`es`	`eswiki_content`	1,276,291	16
Tamil	`tamil`	`ta`	`tawiki_content`	158,025	1
Thai	`thai`	`th`	`thwiki_content`	82,938	2
Vietnamese	`vietnamese`	`vi`	`viwiki_content`	1,129,664	4

Columns

page_id, title, url
text: Cirrus plain text for retrieval and chunking
opening_text: lead/opening text when available
headings, categories, auxiliary_text
wiki, language, wikimedia_language, language_name
timestamp, version, wikibase_item
source_dump_date, source_index_name, source_url, shard_index

Source and license

Source root: https://dumps.wikimedia.org/other/cirrus_search_index/20260628/.

Wikipedia content is available under CC BY-SA/GFDL terms. Downstream training releases must preserve attribution/share-alike obligations where applicable.

Build

Generated on Modal. Large dump files were streamed directly from Wikimedia and were not stored on a local machine.