Datasets:

oscar-corpus
/

community-oscar

Name: community-oscar
Creator: OSCAR
License: https://choosealicense.com/licenses/cc0-1.0/

Tasks:

Sub-tasks:

Size:

ArXiv:

License:

Dataset card Viewer Files Files and versions Community

You need to agree to share your contact information to access this dataset

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

By filling the form below I understand that Community-OSCAR is just a partial annotation of the WET files of 41 Common Crawl snapshots, the original data is included here only for convenience, and specially for researchers looking for data in lower resource languages. Only the annotations are distributed under a cc0-1.0 license, for the rest of the content I have read the Common Crawl Terms of use and I will abide by them. I understand that all uses of the textual content in Community-OSCAR are subject to the Common Crawl Terms of use. I understand that reusing the textual content in Community-OSCAR might not be legal in all countries/regions and for all use cases. I understand that Community-OSCAR is mainly targeted towards researchers and meant to be used in research. The OSCAR Project reserves the right to revoke my access to this data. The OSCAR Project reserves the right to modify this data at any time in accordance to take down requests.

Community OSCAR

The OSCAR project (Open Super-large Crawled Aggregated coRpus) is an Open Source project aiming to provide web-based multilingual resources and datasets for Machine Learning (ML) and Artificial Intelligence (AI) applications. The project focuses specifically in providing large quantities of unannotated raw data that is commonly used in the pre-training of large deep learning models. The OSCAR project has developed high-performance data pipelines specifically conceived to classify and filter large amounts of web data. The project has also put special attention in improving the data quality of web-based corpora as well as providing data for low-resource languages, so that these new ML/AI technologies are accessible to as many communities as possible.

Community-OSCAR is an unofficial version of the OSCAR Corpus created by community members. The annotation schema follows the OSCAR 23.01 release but is based on 41 monthly dumps of Common Crawl ranging from 2024-22 to 2014-42. With these forty dumps, Community-OSCAR is the largest release of the OSCAR Corpus so far.

The annotations contain features, including KenLM-based adult content detection, precomputed Locality-Sensitive Hashes for near deduplication, and blocklist-based categories. Community-OSCAR is distributed as JSONL-files with Zstandard compression. You might already have zstd installed on your system, but if not, please check the Zstandard website for installation instructions.

Ongoing Community Effort

Community-OSCAR is a release created by members of the OSCAR community and part of an ongoing effort in close collaboration with the Occiglot research collective. We are working on extending this release to all publicly available Common Crawl dumps, producing a filtered version (see Occiglot-FineWeb), and have plenty of ideas on further improvements. If you want to support our activities and collaborate with us, please join the Discord server from the OSCAR project or Occiglot research collective.

Downloading the Data

You can stream the data directly into datasets using a script like this

from datasets import load_dataset
# Load afrkaans subset from 2024-22 snapshot
ds = load_dataset('oscar-corpus/community-oscar', data_files='data/2024-22/af_meta/*.jsonl.zst', split='train',
                  streaming=True)

Alternatively you can download the data using huggingface_hub python library.

If you want to download a considerable amount of data we recomend you use hf_transfer python package and set the environment variable HF_HUB_ENABLE_HF_TRANSFER=1.

Supported Tasks and Leaderboards

OSCAR is mainly intended to pre-train language models and word representations.

NOTE: Community-OSCAR contains the raw unfiltered Common Crawl text data but with quality annotations. For language model training, we highly recommend filtering the data first with these annotations. A prefiltered version of the dataset will be released in the near future (following the approach from Occiglot-FineWeb).

Data Annotations

Each sample comes with a series of annotations that allow the removal of low quality data.

identification: Language identification based on fastText.
harmful_pp: This perplexity comes from a KenLM model trained on harmful content, previously gathered by using the adult annotation in OSCAR 22.01. In other terms, the lower it is, the more likely a given document contains harmful/adult content.
tlsh: We use TLSH to compute a hash for each document. Locality sensitive hashing is a hashing method that computes similar hashes for similar documents.
quality_warnings: Computed through heuristics (see below).
categories: Content categories arom a URL-based blocklist

The annotation schema is the same as in the OSCAR 23.01 release.

Quality Warnings

tiny: The document has a low (<5) number of lines.
short_sentences: The document has a high number (>50%) of short lines (<400 bytes)
header: The document has a high number of short lines at its head, suggesting the presence of low quality content.
footer: The document has a high number of short lines at its tail, suggesting the presence of low quality content.
noisy: The document has a high percentage of punctuation (>50%)
adult: The document contains adult content. This annotation uses a blocklist and labels a tiny part of the corpus: It does not catch most of the adult content.

More information about the thresholds and annotators are present in the OSCAR paper.

Data Format

The data is stored as ZSTD-compressed JSON line files. Each individual data sample has the following JSON schema:

{
   "content":"English sentence\nphrase en français\n????????????", // (1)
   "warc_headers":{ // (2)
      "warc-identified-content-language":"fra,eng",
      "warc-target-uri":"https://fr.wikipedia.org/wiki/...",
      "warc-record-id":"<urn:uuid:29eaa920-d299-4b1d-b687-c72bd8d68116>",
      "warc-type":"conversion",
      "content-length":"35298", // (3)
      "warc-refers-to":"<urn:uuid:39e42055-0d94-4e45-9c6c-9e7056635d64>",
      "warc-block-digest":"sha1:WFH2A5WHCS2H365GIAFYQPI7UOAMFGHB", // (3)
      "warc-date":"2022-11-26T09:45:47Z",
      "content-type":"text/plain"
   },
   "metadata":{
      "identification":{ // (4)
         "label":"fr",
         "prob":0.8938327
      },
      "harmful_pp":4063.1814, // (5)
      "tlsh":"tlsh:T125315FF2B6088901EEA097015DB39B4600B...", // (6)
      "quality_warnings":[ // (7)
         "short_sentences",
         "header",
         "footer"
      ],
      "categories":[ // (8)
         "examen_pix",
         "liste_bu"
      ],
      "sentence_identifications":[ // (9)
         {
            "label":"fr",
            "prob":0.99837273
         },
         {
            "label":"en",
            "prob":0.9992377
         },
         null
      ]
   }
}

Language Statistics

All the data is distributed by language and release name. Up to 151 different languages are available. The table below provides the language code as well as the number of documents, compressed data sizes, number of words (by whitespaces) and characters. The statistics are computed based on uncompressed data and on estimates calculated on a subset of 12 releases and extrapolated to all 41 releases (snapshots). Roughly half of the data is English, leaving 19T non-english words.

lang	Language	Data (avg./release)	#Docs (avg./release)	#Words (avg./release)	#Characters (avg./release)	Data (Total)	Lines (Total)	Words (Total)	Characters (Total)
Total		6.78TiB	1.13B	935.22B	6.12T	278.01TiB	46.46B	38.34T	251.03T
en	English	2.56TiB	483.20M	465.20B	2.80T	105.16TiB	19.81B	19.07T	114.75T
Total (w/o en)		4.22TiBB	650.05M	470.02B	3.32T	172.85TiB	26.65B	19.27T	136.27T
ru	Russian	982.11GiB	94.21M	80.68B	593.06B	39.32TiB	3.86B	3.31T	24.32T
zh	Chinese	590.68GiB	67.47M	15.71B	242.67B	23.65TiB	2.77B	644.28B	9.95T
de	German	405.00GiB	79.64M	58.54B	427.68B	16.22TiB	3.27B	2.40T	17.53T
es	Spanish	349.22GiB	58.54M	59.59B	367.67B	13.98TiB	2.40B	2.44T	15.07T
fr	French	306.61GiB	57.83M	50.73B	317.41B	12.28TiB	2.37B	2.08T	13.01T
it	Italian	184.18GiB	31.41M	29.58B	195.26B	7.37TiB	1.29B	1.21T	8.01T
ja	Japanese	165.97GiB	42.67M	5.72B	72.52B	6.65TiB	1.75B	234.47B	2.97T
pt	Portuguese	151.63GiB	26.79M	25.09B	157.95B	6.07TiB	1.10B	1.03T	6.48T
pl	Polish	123.65GiB	22.79M	17.70B	126.32B	4.95TiB	934.43M	725.54B	5.18T
nl	Dutch	91.28GiB	24.44M	15.17B	97.54B	3.65TiB	1.00B	621.77B	4.00T
vi	Vietnamese	83.07GiB	11.49M	14.80B	69.26B	3.33TiB	471.02M	606.86B	2.84T
th	Thai	69.61GiB	5.39M	2.50B	29.73B	2.79TiB	220.82M	102.43B	1.22T
el	Greek	68.44GiB	7.33M	6.24B	42.67B	2.74TiB	300.68M	255.91B	1.75T
tr	Turkish	68.31GiB	13.11M	8.95B	67.37B	2.74TiB	537.64M	366.80B	2.76T
fa	Persian	66.77GiB	8.60M	7.76B	40.94B	2.67TiB	352.65M	318.20B	1.68T
ar	Arabic	61.53GiB	8.57M	6.28B	37.83B	2.46TiB	351.41M	257.37B	1.55T
cs	Czech	54.99GiB	12.97M	8.07B	53.42B	2.20TiB	531.97M	330.77B	2.19T
hu	Hungarian	44.35GiB	7.43M	5.80B	43.57B	1.78TiB	304.66M	237.74B	1.79T
uk	Ukrainian	44.22GiB	5.21M	3.65B	26.80B	1.77TiB	213.63M	149.61B	1.10T
sv	Swedish	42.76GiB	9.03M	6.95B	43.99B	1.71TiB	370.36M	284.99B	1.80T
bg	Bulgarian	33.90GiB	3.49M	3.21B	20.78B	1.36TiB	142.90M	131.57B	852.02B
ro	Romanian	33.50GiB	4.65M	5.29B	34.25B	1.34TiB	190.83M	216.89B	1.40T
ko	Korean	32.35GiB	6.64M	3.31B	15.73B	1.30TiB	272.25M	135.84B	644.73B
fi	Finnish	29.79GiB	5.48M	3.63B	30.82B	1.19TiB	224.50M	148.72B	1.26T
he	Hebrew	28.17GiB	3.74M	2.98B	17.41B	1.13TiB	153.25M	122.19B	713.96B
hi	Hindi	21.48GiB	1.77M	1.86B	10.08B	880.81GiB	72.38M	76.09B	413.17B
id	Indonesian	14.67GiB	2.76M	2.25B	15.71B	601.30GiB	113.09M	92.12B	644.31B
lt	Lithuanian	12.72GiB	2.36M	1.68B	12.83B	521.35GiB	96.72M	69.01B	526.16B
bn	Bangla	12.68GiB	1.26M	834.94M	5.49B	520.02GiB	51.81M	34.23B	225.10B
sk	Slovak	12.19GiB	2.85M	1.77B	12.03B	499.80GiB	116.77M	72.49B	493.20B
da	Danish	11.38GiB	2.87M	1.93B	11.90B	466.69GiB	117.50M	79.22B	488.04B
ca	Catalan	11.16GiB	2.81M	1.86B	11.63B	457.38GiB	115.40M	76.42B	476.63B
ta	Tamil	10.33GiB	574763	559.30M	4.59B	423.46GiB	23.57M	22.93B	188.20B
multi	-	8.85GiB	1.22M	1.05B	7.14B	362.94GiB	49.85M	43.22B	292.56B
ka	Georgian	7.31GiB	592883	387.90M	3.14B	299.82GiB	24.31M	15.90B	128.83B
et	Estonian	7.24GiB	1.59M	976.06M	7.53B	296.66GiB	65.37M	40.02B	308.78B
sr	Serbian	7.19GiB	704978	675.30M	4.44B	294.65GiB	28.90M	27.69B	182.02B
lv	Latvian	7.12GiB	1.21M	936.00M	7.03B	291.95GiB	49.69M	38.38B	288.11B
hy	Armenian	4.43GiB	424969	348.89M	2.70B	181.46GiB	17.42M	14.30B	110.56B
ml	Malayalam	4.32GiB	291610	198.34M	1.85B	177.15GiB	11.96M	8.13B	75.82B
az	Azerbaijani	3.36GiB	603832	408.83M	3.12B	137.62GiB	24.76M	16.76B	128.07B
te	Telugu	3.34GiB	267406	194.03M	1.49B	136.82GiB	10.96M	7.96B	60.95B
kk	Kazakh	3.24GiB	320143	239.00M	1.93B	132.95GiB	13.13M	9.80B	79.00B
ne	Nepali	3.17GiB	400031	204.06M	1.33B	130.11GiB	16.40M	8.37B	54.35B
mr	Marathi	3.00GiB	251313	194.56M	1.32B	122.82GiB	10.30M	7.98B	54.26B
ur	Urdu	2.66GiB	342973	342.83M	1.66B	109.20GiB	14.06M	14.06B	68.00B
mk	Macedonian	2.57GiB	379079	243.56M	1.58B	105.49GiB	15.54M	9.99B	64.72B
sq	Albanian	2.48GiB	496860	407.12M	2.48B	101.56GiB	20.37M	16.69B	101.67B
my	Burmese	2.36GiB	173587	79.12M	968.68M	96.60GiB	7.12M	3.24B	39.72B
gu	Gujarati	2.31GiB	128983	194.10M	1.15B	94.72GiB	5.29M	7.96B	47.23B
kn	Kannada	2.04GiB	153938	117.35M	951.65M	83.81GiB	6.31M	4.81B	39.02B
be	Belarusian	2.00GiB	234797	163.91M	1.21B	82.05GiB	9.63M	6.72B	49.51B
no	Norwegian	1.97GiB	1.11M	320.46M	2.07B	80.65GiB	45.70M	13.14B	84.72B
is	Icelandic	1.89GiB	479673	285.65M	1.83B	77.44GiB	19.67M	11.71B	75.23B
mn	Mongolian	1.87GiB	213630	162.46M	1.15B	76.86GiB	8.76M	6.66B	47.35B
km	Khmer	1.79GiB	140645	40.87M	714.22M	73.31GiB	5.77M	1.68B	29.28B
si	Sinhala	1.78GiB	112624	134.15M	844.35M	73.06GiB	4.62M	5.50B	34.62B
sl	Slovenian	1.01GiB	445779	156.82M	1.06B	41.59GiB	18.28M	6.43B	43.45B
tg	Tajik	988.00MiB	73726	83.29M	576.25M	39.56GiB	3.02M	3.41B	23.63B
eu	Basque	806.16MiB	262487	104.16M	842.18M	32.28GiB	10.76M	4.27B	34.53B
pa	Punjabi	780.76MiB	71240	65.56M	350.37M	31.26GiB	2.92M	2.69B	14.36B
tt	Tatar	684.17MiB	77931	56.24M	402.50M	27.39GiB	3.20M	2.31B	16.50B
ckb	Central Kurdish	624.12MiB	93207	52.42M	360.95M	24.99GiB	3.82M	2.15B	14.80B
ky	Kyrgyz	489.66MiB	72858	36.24M	286.59M	19.61GiB	2.99M	1.49B	11.75B
tl	Filipino	460.45MiB	74172	77.36M	480.16M	18.44GiB	3.04M	3.17B	19.69B
am	Amharic	436.81MiB	44448	38.29M	205.38M	17.49GiB	1.82M	1.57B	8.42B
eo	Esperanto	411.92MiB	107893	64.90M	421.28M	16.49GiB	4.42M	2.66B	17.27B
or	Odia	364.42MiB	54702	24.03M	157.21M	14.59GiB	2.24M	985.22M	6.45B
bo	Tibetan	335.38MiB	22403	4.47M	126.55M	13.43GiB	918536	183.15M	5.19B
ps	Pashto	293.88MiB	45796	37.31M	178.25M	11.77GiB	1.88M	1.53B	7.31B
lo	Lao	286.92MiB	35277	7.46M	113.82M	11.49GiB	1.45M	305.88M	4.67B
cy	Welsh	268.65MiB	78645	46.25M	278.10M	10.76GiB	3.22M	1.90B	11.40B
ug	Uyghur	212.29MiB	22044	15.59M	121.98M	8.50GiB	903834	639.20M	5.00B
dv	Divehi	207.92MiB	30485	13.36M	119.18M	8.33GiB	1.25M	547.74M	4.89B
as	Assamese	202.40MiB	18951	13.49M	88.34M	8.10GiB	776997	552.99M	3.62B
gl	Galician	195.43MiB	96533	31.92M	199.39M	7.82GiB	3.96M	1.31B	8.18B
yi	Yiddish	157.11MiB	20405	15.33M	94.83M	6.29GiB	836632	628.66M	3.89B
ba	Bashkir	155.53MiB	23672	12.54M	92.19M	6.23GiB	970562	514.29M	3.78B
ku	Kurdish	120.61MiB	31901	19.63M	114.28M	4.83GiB	1.31M	805.03M	4.69B
sd	Sindhi	110.46MiB	14589	13.74M	67.60M	4.42GiB	598183	563.48M	2.77B
hr	Croatian	88.36MiB	14480	12.35M	88.91M	3.54GiB	593680	506.36M	3.65B
sa	Sanskrit	87.88MiB	7776	4.39M	34.88M	3.52GiB	318843	180.14M	1.43B
pnb	Western Panjabi	53.99MiB	8182	6.47M	32.91M	2.16GiB	335482	265.46M	1.35B
sah	Yakut	50.51MiB	7804	3.55M	29.19M	2.02GiB	319984	145.57M	1.20B
fy	Western Frisian	50.19MiB	23703	7.91M	50.53M	2.01GiB	971826	324.19M	2.07B
cv	Chuvash	41.20MiB	6514	3.40M	24.15M	1.65GiB	267111	139.23M	990.10M
ga	Irish	33.83MiB	15186	5.41M	33.13M	1.35GiB	622656	221.81M	1.36B
ceb	Cebuano	30.24MiB	4924	4.79M	31.23M	1.21GiB	201918	196.19M	1.28B
af	Afrikaans	27.95MiB	12458	5.07M	28.89M	1.12GiB	510784	207.74M	1.18B
br	Breton	26.95MiB	22426	4.53M	27.41M	1.08GiB	919479	185.89M	1.12B
os	Ossetic	19.94MiB	6503	1.70M	11.90M	817.46MiB	266636	69.59M	488.10M
uz	Uzbek	16.15MiB	13478	1.97M	16.45M	662.13MiB	552625	80.91M	674.62M
azb	South Azerbaijani	14.83MiB	7798	1.19M	8.69M	607.84MiB	319721	48.88M	356.10M
lb	Luxembourgish	13.08MiB	7289	1.96M	13.28M	536.38MiB	298849	80.38M	544.40M
mg	Malagasy	12.88MiB	3983	1.85M	13.45M	527.98MiB	163309	76.00M	551.43M
mhr	Eastern Mari	10.49MiB	2345	830225	6.13M	430.25MiB	96145	34.04M	251.16M
nds	Low German	9.58MiB	2046	1.58M	9.71M	392.94MiB	83916	64.77M	398.20M
ce	Chechen	8.88MiB	3313	735487	5.22M	363.94MiB	135870	30.15M	214.08M
xmf	Mingrelian	7.23MiB	2959	393400	3.15M	296.34MiB	121322	16.13M	129.24M
new	Newari	5.02MiB	916	324361	2.08M	205.73MiB	37569	13.30M	85.38M
sh	Serbian	4.22MiB	1019	1.02M	4.28M	173.18MiB	41779	41.69M	175.64M
ms	Malay	4.02MiB	4988	382700	3.32M	164.63MiB	204511	15.69M	136.14M
min	Minangkabau	3.63MiB	953	324454	2.07M	148.63MiB	39103	13.30M	84.97M
nn	Norwegian Nynorsk	3.30MiB	8285	553667	3.37M	135.40MiB	339685	22.70M	138.31M
tk	Turkmen	2.41MiB	1662	269719	2.28M	98.64MiB	68162	11.06M	93.31M
gom	Goan Konkani	2.14MiB	135	127985	891404	87.80MiB	5541	5.25M	36.55M
arz	Egyptian Arabic	2.11MiB	1482	234425	1.26M	86.36MiB	60789	9.61M	51.84M
bpy	Bishnupriya	1.95MiB	379	130226	824058	80.09MiB	15566	5.34M	33.79M
la	Latin	1.61MiB	4738	276095	1.68M	66.12MiB	194281	11.32M	69.02M
pms	Piedmontese	1.49MiB	466	264178	1.47M	61.22MiB	19112	10.83M	60.40M
jbo	Lojban	1.30MiB	276	277677	1.35M	53.38MiB	11329	11.38M	55.40M
mt	Maltese	1.17MiB	2544	145437	1.17M	48.00MiB	104324	5.96M	48.04M
oc	Occitan	1020.71KiB	280	38821	1.03M	40.87MiB	11510	1.59M	42.35M
war	Waray	997.57KiB	328	150252	1.02M	39.94MiB	13468	6.16M	41.73M
vo	Volapük	935.64KiB	572	140394	888363	37.46MiB	23482	5.76M	36.42M
ast	Asturian	599.07KiB	1742	90259	592203	23.99MiB	71459	3.70M	24.28M
lez	Lezghian	454.96KiB	145	35011	259634	18.22MiB	5962	1.44M	10.65M
mrj	Western Mari	419.70KiB	130	33093	242203	16.80MiB	5340	1.36M	9.93M
su	Sundanese	394.30KiB	35	61531	388935	15.79MiB	1438	2.52M	15.95M
sw	Swahili	376.35KiB	683	64274	384592	15.07MiB	28033	2.64M	15.77M
gsw	Swiss German	367.62KiB	163	54335	352731	14.72MiB	6700	2.23M	14.46M
wuu	Wu Chinese	232.88KiB	86	6765	86236	9.32MiB	3532	277388	3.54M
wa	Walloon	171.79KiB	52	3278	175434	6.88MiB	2132	134435	7.19M
gd	Scottish Gaelic	91.63KiB	253	10844	91821	3.67MiB	10403	444624	3.76M
mzn	Mazanderani	88.55KiB	50	9282	50922	3.55MiB	2080	380582	2.09M
hsb	Upper Sorbian	85.59KiB	134	11768	80704	3.43MiB	5504	482505	3.31M
ia	Interlingua	80.21KiB	37	23253	81382	3.21MiB	1527	953407	3.34M
krc	Karachay-Balkar	58.04KiB	83	4072	32574	2.32MiB	3420	166969	1.34M
kv	Komi	25.03KiB	67	2182	14450	1.00MiB	2760	89479	592456
av	Avaric	23.69KiB	27	1440	13157	971.31KiB	1113	59053	539450
jv	Javanese	18.04KiB	46	2534	17534	739.83KiB	1886	103928	718917
ilo	Iloko	15.43KiB	39	2554	15737	632.47KiB	1599	104737	645234
li	Limburgish	13.46KiB	2	77	13762	552.02KiB	118	3161	564260
mai	Maithili	11.20KiB	18	1455	5092	459.00KiB	738	59689	208806
yo	Yoruba	10.89KiB	43	1502	7585	446.46KiB	1773	61595	311008
lmo	Lombard	9.66KiB	26	1582	9200	396.21KiB	1066	64896	377227
an	Aragonese	7.55KiB	11	264	4311	309.74KiB	481	10851	176751
bar	Bavarian	6.82KiB	26	1743	3543	279.77KiB	1090	71479	145287
io	Ido	6.51KiB	43	1114	6633	266.79KiB	1797	45674	271970
bh	Bihari languages	5.90KiB	21	465	2325	241.72KiB	866	19065	95354
bxr	Russia Buriat	5.16KiB	23	470	3036	211.74KiB	951	19278	124476
bs	Bosnian	3.85KiB	8	474	3802	157.87KiB	328	19464	155912
ie	Interlingue	3.06KiB	1	722	2967	125.60KiB	61	29602	121647
so	Somali	2.26KiB	18	647	2114	92.64KiB	738	26556	86692
xal	Kalmyk	2.02KiB	6	189	1184	82.73KiB	280	7769	48544
nah	Nahuatl languages	1.80KiB	13	192	1690	74.00KiB	553	7884	69290
gn	Guarani	1.51KiB	5	191	1375	61.90KiB	222	7836	56386
ht	Haitian Creole	85.38B	1	118	681	27.35KiB	41	4858	27921
kw	Cornish	70.14B	3	104	559	22.47KiB	127	4272	22931
x-eml	Unknown language [x-eml]	61.71B	1	85	445	19.77KiB	41	3485	18258
lrc	Northern Luri	52.33B	1	41	232	16.76KiB	68	1681	9512
dsb	Lower Sorbian	36.08B	1	37	260	11.56KiB	68	1517	10687
rue	Rusyn	29.75B	1	8	130	9.53KiB	41	328	5330
scn	Sicilian	27.12B	1	37	204	8.69KiB	49	1525	8380
qu	Quechua	26.51B	1	24	207	8.49KiB	59	997	8487
vec	Venetian	20.00B	1	34	151	6.41KiB	41	1394	6191
diq	Dimli (individual language)	18.46B	1	19	132	5.91KiB	41	779	5425
rm	Romansh	13.00B	1	10	104	4.16KiB	41	410	4264
	-	0.00B	0	0	0	0.00B	0	0	0

Issues

Community-OSCAR may have quality issues on low size subcorpora, as it has been the case before. Please consider taking a look at Quality at a Glance: An Audit of Web-Crawled Multilingual Datasets to get a better understanding of the current limitations of our language classifier.

Note that since the documents are identified as a whole, it is expected to have lines in other languages in a given language subcorpus. As an example, it is known and expected that the German subcorpus contains documents holding lines identified as Swiss German / Alemannic.

If you encounter something that is unexpected, please file an issue here: https://github.com/oscar-corpus/corpus/issues.

Dataset Creation

Curation Rationale

OSCAR was constructed using Ungoliant, a new pipeline derived from goclassy, itself being derived from fastText's one.

The pipeline works on documents rather than lines. Ungoliant is implemented in the Rust programming language, and uses rayon as its data parallelism strategy. Threading is done at shard, record and sentence level, making the whole generation process much more efficient.

Filtering will be explained in a future blog post at our website

Source Data

Common Crawl is a non-profit foundation which produces and maintains an open repository of web crawled data that is both accessible and analysable. Common Crawl's complete web archive consists of petabytes of data collected over 8 years of web crawling. The repository contains raw web page HTML data (WARC files), metdata extracts (WAT files) and plain text extracts (WET files). The organization's crawlers has always respected nofollow and robots.txt policies.

Each monthly Common Crawl snapshot is in itself a massive multilingual corpus, where every single file contains data coming from multiple web pages written in a large variety of languages and covering all possible types of topics.

To construct OSCAR the WET files of Common Crawl were used. These contain the extracted plain texts from the websites mostly converted to UTF-8, as well as headers containing the metatada of each crawled document. Each WET file comes compressed in gzip format and is stored on Amazon Web Services.

In the case of Community-OSCAR the following forty monthly Common Crawl snapshots were used:

Who are the source language producers?

The data comes from multiple web pages in a large variety of languages.

Personal and Sensitive Information

Being constructed from Common Crawl, Personal and sensitive information might be present. This must be considered before training deep learning models with OSCAR, specially in the case of text-generation models.

Considerations for Using the Data

Social Impact of Dataset

OSCAR is intended to bring more data to a wide variety of languages, the aim of the corpus is to make large amounts of data available to lower resource languages in order to facilitate the pre-training of state-of-the-art language modeling architectures.

Discussion of Biases

OSCAR is not properly filtered yet and this can be reflected on the models trained with it. Care is advised specially concerning biases of the resulting models. We have added annotations to Common Crawl, so please consider using them to select the data that you would like to use for your particular use case.

Other Known Limitations

The fastText linear classifier is limed both in performance and the variety of languages it can recognize, so the quality of some OSCAR sub-corpora might be lower than expected, specially for the lowest-resource languages. Some audits have already been done by third parties.

Dataset Curators & Contributors

Community-OSCAR was put together by community members in close collaboration with the Occiglot research collective. The main contributors are Manuel Brack, Pedro Ortiz Suarez, Malte Ostendorff, Patrick Schramowski, Georg Rehm, Kristian Kersting, Jose Javier Saiz, Iñaki Lacunza Castilla, Alexander Shvets, Jorge Palomar-Giner, and Marta Villegas. Moreover, this release is supported by and was enabled by contributions from the OSCAR team at Inria (project-team ALMAnaCH), specially by Julien Abadji, Rua Ismail and Benoit Sagot, the Common Crawl Foundation, the SLT and SAINT teams at DFKI, TU Darmstadt, the LangTech unit at the Barcelona Supercomputing Center, the 42 supercomputer and Hessian AI, the OpenGPT-X project, Fraunhofer, Jülich Supercomputing Centre, TU Dresden, Deutsche Telekom, as well as by members of the OSCAR community, in particular Sotaro Takeshita, Sebastian Nagel.

Licensing Information

These data are released under this licensing scheme We do not own any of the text from which these data has been extracted. We license the actual packaging, the metadata and the annotations of these data under the Creative Commons CC0 license ("no rights reserved") http://creativecommons.org/publicdomain/zero/1.0/ To the extent possible under law, the OSCAR community members have waived all copyright and related or neighboring rights to OSCAR.

Should you consider that our data contains material that is owned by you and should therefore not be reproduced here, please:

Clearly identify yourself, with detailed contact data such as an address, telephone number or email address at which you can be contacted.
Clearly identify the copyrighted work claimed to be infringed.
Clearly identify the material that is claimed to be infringing and information reasonably sufficient to allow us to locate the material.

We will comply to legitimate requests by removing the affected sources.

Please use the contact information on our website for take down requests.

We strongly advise users to submit take down request to Common Crawl. For more information please read their Terms of Use

Citation Information

If you use our work, please cite the technical report:

@misc{brack2024communityoscar,
    title        = {Community OSCAR: A Community Effort for Multilingual Web Data},
    author       = {Manuel Brack and Malte Ostendorff and Pedro Ortiz Suarez and José Javier Saiz and Iñaki Lacunza Castilla and Jorge Palomar-Giner and Aleksandr Shvets and Patrick Schramowski and Georg Rehm and Marta Villegas and Kristian Kersting},
    year         = {2024},
    howpublished = {technical report},
    url          = {https://occiglot.eu/papers/Community_Oscar.pdf}
}

Additionally, please consider these citations that Community OSCAR relies on:

@ARTICLE{2022arXiv221210440J,
       author = {{Jansen}, Tim and {Tong}, Yangling and {Zevallos}, Victoria and {Ortiz Suarez}, Pedro},
        title = "{Perplexed by Quality: A Perplexity-based Method for Adult and Harmful Content Detection in Multilingual Heterogeneous Web Data}",
      journal = {arXiv e-prints},
     keywords = {Computer Science - Computation and Language},
         year = 2022,
        month = dec,
          eid = {arXiv:2212.10440},
        pages = {arXiv:2212.10440},
          doi = {10.48550/arXiv.2212.10440},
archivePrefix = {arXiv},
       eprint = {2212.10440},
 primaryClass = {cs.CL},
       adsurl = {https://ui.adsabs.harvard.edu/abs/2022arXiv221210440J},
      adsnote = {Provided by the SAO/NASA Astrophysics Data System}
}

@inproceedings{abadji-etal-2022-towards,
    title = "Towards a Cleaner Document-Oriented Multilingual Crawled Corpus",
    author = "Abadji, Julien  and
      Ortiz Suarez, Pedro  and
      Romary, Laurent  and
      Sagot, Beno{\^\i}t",
    booktitle = "Proceedings of the Thirteenth Language Resources and Evaluation Conference",
    month = jun,
    year = "2022",
    address = "Marseille, France",
    publisher = "European Language Resources Association",
    url = "https://aclanthology.org/2022.lrec-1.463",
    pages = "4344--4355",
    abstract = "The need for large corpora raw corpora has dramatically increased in recent years with the introduction of transfer learning and semi-supervised learning methods to Natural Language Processing. And while there have been some recent attempts to manually curate the amount of data necessary to train large language models, the main way to obtain this data is still through automatic web crawling. In this paper we take the existing multilingual web corpus OSCAR and its pipeline Ungoliant that extracts and classifies data from Common Crawl at the line level, and propose a set of improvements and automatic annotations in order to produce a new document-oriented version of OSCAR that could prove more suitable to pre-train large generative language models as well as hopefully other applications in Natural Language Processing and Digital Humanities.",
}

      
@inproceedings{AbadjiOrtizSuarezRomaryetal.2021,
  author    = {Julien Abadji and Pedro Javier Ortiz Su{\'a}rez and Laurent Romary and Beno{\^i}t Sagot},
  title     = {Ungoliant: An optimized pipeline for the generation of a very large-scale multilingual web corpus},
  series = {Proceedings of the Workshop on Challenges in the Management of Large Corpora (CMLC-9) 2021. Limerick, 12 July 2021 (Online-Event)},
  editor    = {Harald L{\"u}ngen and Marc Kupietz and Piotr Bański and Adrien Barbaresi and Simon Clematide and Ines Pisetta},
  publisher = {Leibniz-Institut f{\"u}r Deutsche Sprache},
  address   = {Mannheim},
  doi       = {10.14618/ids-pub-10468},
  url       = {https://nbn-resolving.org/urn:nbn:de:bsz:mh39-104688},
  pages     = {1 -- 9},
  year      = {2021},
  abstract  = {Since the introduction of large language models in Natural Language Processing, large raw corpora have played a crucial role in Computational Linguistics. However, most of these large raw corpora are either available only for English or not available to the general public due to copyright issues. Nevertheless, there are some examples of freely available multilingual corpora for training Deep Learning NLP models, such as the OSCAR and Paracrawl corpora. However, they have quality issues, especially for low-resource languages. Moreover, recreating or updating these corpora is very complex. In this work, we try to reproduce and improve the goclassy pipeline used to create the OSCAR corpus. We propose a new pipeline that is faster, modular, parameterizable, and well documented. We use it to create a corpus similar to OSCAR but larger and based on recent data. Also, unlike OSCAR, the metadata information is at the document level. We release our pipeline under an open source license and publish the corpus under a research-only license.},
  language  = {en}
}

@article{kreutzer-etal-2022-quality,
    title = "Quality at a Glance: An Audit of Web-Crawled Multilingual Datasets",
    author = {Kreutzer, Julia  and
      Caswell, Isaac  and
      Wang, Lisa  and
      Wahab, Ahsan  and
      van Esch, Daan  and
      Ulzii-Orshikh, Nasanbayar  and
      Tapo, Allahsera  and
      Subramani, Nishant  and
      Sokolov, Artem  and
      Sikasote, Claytone  and
      Setyawan, Monang  and
      Sarin, Supheakmungkol  and
      Samb, Sokhar  and
      Sagot, Beno{\^\i}t  and
      Rivera, Clara  and
      Rios, Annette  and
      Papadimitriou, Isabel  and
      Osei, Salomey  and
      Suarez, Pedro Ortiz  and
      Orife, Iroro  and
      Ogueji, Kelechi  and
      Rubungo, Andre Niyongabo  and
      Nguyen, Toan Q.  and
      M{\"u}ller, Mathias  and
      M{\"u}ller, Andr{\'e}  and
      Muhammad, Shamsuddeen Hassan  and
      Muhammad, Nanda  and
      Mnyakeni, Ayanda  and
      Mirzakhalov, Jamshidbek  and
      Matangira, Tapiwanashe  and
      Leong, Colin  and
      Lawson, Nze  and
      Kudugunta, Sneha  and
      Jernite, Yacine  and
      Jenny, Mathias  and
      Firat, Orhan  and
      Dossou, Bonaventure F. P.  and
      Dlamini, Sakhile  and
      de Silva, Nisansa  and
      {\c{C}}abuk Ball{\i}, Sakine  and
      Biderman, Stella  and
      Battisti, Alessia  and
      Baruwa, Ahmed  and
      Bapna, Ankur  and
      Baljekar, Pallavi  and
      Azime, Israel Abebe  and
      Awokoya, Ayodele  and
      Ataman, Duygu  and
      Ahia, Orevaoghene  and
      Ahia, Oghenefego  and
      Agrawal, Sweta  and
      Adeyemi, Mofetoluwa},
    journal = "Transactions of the Association for Computational Linguistics",
    volume = "10",
    year = "2022",
    address = "Cambridge, MA",
    publisher = "MIT Press",
    url = "https://aclanthology.org/2022.tacl-1.4",
    doi = "10.1162/tacl_a_00447",
    pages = "50--72",
    abstract = "With the success of large-scale pre-training and multilingual modeling in Natural Language Processing (NLP), recent years have seen a proliferation of large, Web-mined text datasets covering hundreds of languages. We manually audit the quality of 205 language-specific corpora released with five major public datasets (CCAligned, ParaCrawl, WikiMatrix, OSCAR, mC4). Lower-resource corpora have systematic issues: At least 15 corpora have no usable text, and a significant fraction contains less than 50{\%} sentences of acceptable quality. In addition, many are mislabeled or use nonstandard/ambiguous language codes. We demonstrate that these issues are easy to detect even for non-proficient speakers, and supplement the human audit with automatic analyses. Finally, we recommend techniques to evaluate and improve multilingual corpora and discuss potential risks that come with low-quality data releases.",
}

@inproceedings{ortiz-suarez-etal-2020-monolingual,
    title = "A Monolingual Approach to Contextualized Word Embeddings for Mid-Resource Languages",
    author = "Ortiz Su{'a}rez, Pedro Javier  and
      Romary, Laurent  and
      Sagot, Benoit",
    booktitle = "Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics",
    month = jul,
    year = "2020",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/2020.acl-main.156",
    pages = "1703--1714",
    abstract = "We use the multilingual OSCAR corpus, extracted from Common Crawl via language classification, filtering and cleaning, to train monolingual contextualized word embeddings (ELMo) for five mid-resource languages. We then compare the performance of OSCAR-based and Wikipedia-based ELMo embeddings for these languages on the part-of-speech tagging and parsing tasks. We show that, despite the noise in the Common-Crawl-based OSCAR data, embeddings trained on OSCAR perform much better than monolingual embeddings trained on Wikipedia. They actually equal or improve the current state of the art in tagging and parsing for all five languages. In particular, they also improve over multilingual Wikipedia-based contextual embeddings (multilingual BERT), which almost always constitutes the previous state of the art, thereby showing that the benefit of a larger, more diverse corpus surpasses the cross-lingual benefit of multilingual embedding architectures.",
}

@inproceedings{OrtizSuarezSagotRomary2019,
  author    = {Pedro Javier {Ortiz Su{'a}rez} and Benoit Sagot and Laurent Romary},
  title     = {Asynchronous pipelines for processing huge corpora on medium to low resource infrastructures},
  series = {Proceedings of the Workshop on Challenges in the Management of Large Corpora (CMLC-7) 2019. Cardiff, 22nd July 2019},
  editor    = {Piotr Bański and Adrien Barbaresi and Hanno Biber and Evelyn Breiteneder and Simon Clematide and Marc Kupietz and Harald L{"u}ngen and Caroline Iliadi},
  publisher = {Leibniz-Institut f{"u}r Deutsche Sprache},
  address   = {Mannheim},
  doi       = {10.14618/ids-pub-9021},
  url       = {http://nbn-resolving.de/urn:nbn:de:bsz:mh39-90215},
  pages     = {9 -- 16},
  year      = {2019},
  abstract  = {Common Crawl is a considerably large, heterogeneous multilingual corpus comprised of crawled documents from the internet, surpassing 20TB of data and distributed as a set of more than 50 thousand plain text files where each contains many documents written in a wide variety of languages. Even though each document has a metadata block associated to it, this data lacks any information about the language in which each document is written, making it extremely difficult to use Common Crawl for monolingual applications. We propose a general, highly parallel, multithreaded pipeline to clean and classify Common Crawl by language; we specifically design it so that it runs efficiently on medium to low resource infrastructures where I/O speeds are the main constraint. We develop the pipeline so that it can be easily reapplied to any kind of heterogeneous corpus and so that it can be parameterised to a wide range of infrastructures. We also distribute a 6.3TB version of Common Crawl, filtered, classified by language, shuffled at line level in order to avoid copyright issues, and ready to be used for NLP applications.},
  language  = {en}
}

Downloads last month: 41

Edit dataset card Papers with Code