SMOL

SMOL (Set for Maximal Overall Leverage) is a collection professional translations into 221 Low-Resource Languages, for the purpose of training translation models, and otherwise increasing the representations of said languages in NLP and technology.

Please read the SMOL Paper and the GATITOS Paper for a much more thorough description!

There are four resources in this directory:

• SmolDoc: document-level translations into 100 languages
• SmolSent: sentence-level translations into 81 languages
• GATITOS: token-level translations into 172 languages
• SmolDoc-factuality-annotations: factuality annotations and rationales for 661 documents from SmolDoc

Data schemata

The schemata are pretty straightforward. Source and Target languge are provided in sl and tl fields. The is_src_orig field has a value of true if the source text was the original text, and the target field was translated, and avalue of false if the data is back-translated.

GATITOS

GATITOS has a one-to-many schema, with each source mapping to one or more targets:

{
 'sl': 'en',
 'tl': 'aa',
 'is_source_orig': True,
 'src': 'how are you',
 'trgs': ['mannah taniih?', 'anninnaay?']
}

SmolSent

SmolSent's schema differs from SmolDoc's only in that src and trg are single strings, not lists:

{'id': 381,
 'sl': 'en',
 'tl': 'ber',
 'is_src_orig': True,
 'src': 'Rih, a deaf former soldier, plots rebellion while married to a queer, teenage god.',
 'trg': 'ⵔⵉⵀ, ⴷ ⴰⴷⴻⴼⵔⵉⵔ ⴰⵇⴱⵓⵔ ⴰⴻⵎⴻⵙⵍⵉ, ⵢⴻⵜⵜⵀⴻⴳⴳⵉ ⵜⴰⴴⴻⵡⵡⴰⵜ-ⵉⵙ, ⴴⴰⵙ ⴰⴽⴽⴻⵏ ⵢⴻⵣⵡⴻⴵ ⴷ ⵢⵉⵡⴻⵏ ⵢⵉⵍⵓ ⵉⵍⴻⵎⵥⵉ ⵉⵁⴻⵎⵎⵍⴻⵏ ⴰⵔⵔⴰⵛ.'
}

SmolDoc

SmolDoc provides sentence-split and aligned translations through the srcs and trgs fields. These will always have the same length.

{
'id': 'topic_587__weyiwiniwaaotiwenwy',
 'sl': 'en',
 'tl': 'pcm',
 'is_src_orig': True,
 'factuality': 'ok',  # this is a story so there is no factual claim that could be wrong
 'srcs': ['"What the hell are you doing, you idiot?!"',
          '"Excuse me?"', 
          '"You cut me off! You almost made me crash!"', 
          '"I'm sorry, I didn't mean to. I was just trying to get around that slow-moving truck."',
          '"Well, you could have at least used your turn signal!"',
          '"I did use my turn signal!"',
          '"No, you didn't! You just pulled right out in front of me!"',
          '"I'm telling you, I used my turn signal!"',
          '"Whatever. You're still a terrible driver."',
          '"And you're a jerk!"', 
          '"At least I know how to drive!"',
          '"Oh, yeah? Well, I'm a better writer than you are!"',
          '"That's debatable."',
          '"It's not debatable! I'm Ernest Hemingway!"',
          '"Who?"',
          '"Ernest Hemingway! The greatest writer of all time!"',
          '"Never heard of him."',
          '"Well, you've heard of me now!"',
          '"Yeah, I heard of you."'],
 'trgs': ['"Wetin di hell dey do, yu idiot?!"',
          '"Ekskuse mi?"', 
          '"Yu komot mi! Yu almost make mi krash!"',
          '"I dey sorry, I nor wont do am. I just dey try get around dat truk wey slow."',
          '"Well, yu for don yus yor turn sign!"',
          '"I yus mai turn sign!"',
          '"No, yu nor turn am! Yu just turn rite in front of mi!"',
          '"I dey tell yu, I yus mai turn sign!"',
          '"Wateva. Yu still bi one tribol driva."',
          '"And yu bi jerk!"',
          '"At least I sabi hau to drive!"',
          '"Oh, yeah? Well, I bi ogbonge writa pass yu!"',
          '"Wi fit dibate dat."',
          '"nortin to dibate! I bi Ernest Hemingway!"',
          '"Who?"',
          '"Ernest Hemingway! De writa of all taim wey grate pass!"',
          '"Neva hear am."',
          '"Well, yu don hear mi nau!"',
          '"Na so, I don hear yu."']
}

SmolDoc factuality

The factuality map contains the per-document ratings for each of the three raters, for all documents in the SmolDoc data. The source text itself is not included, but can be gotten by joining with SmolDoc on the id field. Example datum:

  "topic_260__mtaftfttit": {
    "1": [
      "Minor Issue(s)",
      "This paragraph is mostly accurate, but there's an issue with the study mentioned. The study by the Geena Davis Institute found that only 36.3% of female characters had speaking roles, not 30% (source: https://geenadavisinstitute.org/wp-content/uploads/2024/01/see-jane-100-report-2017.pdf)."
    ],
    "2": [
      "Minor Issue(s)",
      "The claim regarding 12% of female characters working outside the home is not explicitly supported by Geena Davis Institute studies as seen in the article https://geenadavisinstitute.org/research/gender-bias-with-borders-an-investigation-of-female-characters-in-popular-films-across-11-countriesgender-bias-with-borders/"
    ],
    "3": [
      "Not Sure",
      "The claims are generally true. But I could not confirm that 12% of female characters were working outside the home according to this site. https://geenadavisinstitute.org/wp-content/uploads/2024/01/see-jane-100-report-2017.pdf"
    ]
  },

Notes on the GATITOS multilingual Lexicon

The GATITOS (Google's Additional Translations Into Tail-languages: Often Short) dataset is a high-quality, multi-way parallel dataset of tokens and short phrases, intended for training and improving machine translation models. Experiments on this dataset and Panlex focusing on unsupervised translation in a 208-language model can be found in the GATITOS paper.

This dataset consists in 4,000 English segments (4,500 tokens) that have been translated into each of 173 languages, and as a 4001st token, the endonym. Of these, 170 are low-resource, and three are mid-high resource (es, fr, hi). All translations were made directly from English, with the exception of Aymara, which was translated from the Spanish.

This dataset contains primarily short segments: 93% single tokens, and only 23 sentences (0.6%) have over 5 tokens. As such it is best thought of as a multilingual lexicon, rather than a parallel training corpus. The source text is frequent words in the English Language, along with some common phrases and short sentences. Care has been taken to ensure that they include good coverage of numbers, months, days of the week, swadesh words, and names of the languages themselves (including the endonym).

Single tokens are remarkably fickle to translate because of the common issue of polysemy. Some languages in this dataset only provide one translation per source token; others provide multiple translations. When multiple translations are given, they are broken up into multiple lines. Therefore, you can get the average number of translations per token by looking at the number of lines in the file versus the number of deduplicated English lines in the file. The three languages with the most average translations per English token are Betawi, Kachin, and Occitan, each with over 1.5 on average.

Sometimes translators have left annotations like "pike (verb)", "او (مونث)", "фамили(я)", "أسوأ(repeated)" etc.. Wolof is particularly noticeable in this regard, with notes like "xont (verb) which means to feed an animal". Several languages also give definitions instead of translations for some words, e.g. Tok Pisin translation of mole" to "wanpela liklik animal we i save stap long hol long graun".

Per-language information

The following table only includes information for the source_lang_translated_lang direction. However, data in this repo is released in both directions, so e.g. this table lists only en_ay, but reversed ay_en data is also released.

LP	Target Language Name	ISO 15924 Script	Continent	Endonym	Alternate names	N Smoldoc	N SmolSent	N Gatitos
am_ar	Arabic	Arab	Asia	اللُّغَة العَرَبِيّة	Arabic (MSA); Arabic (Standard); Arabic (Fusha)	329	0	0
am_zh	Mandarin Chinese	Hans	Asia	汉语		329	0	0
en_aa	Afar	Latn	Africa	Qafar		66	863	3990
en_ab	Abkhaz	Cyrl	Asia	Аҧсуа	Abkhazian	0	0	3976
en_ace	Acehnese	Latn	Asia	Aceh	Achinese; Dayan; Aceh	0	0	4001
en_ach	Acholi	Latn	Africa	Acholi	Luo; Acoli	66	863	4001
en_ady	West Circassian	Cyrl	Europe	Черкес	Adyghe	0	0	4001
en_aeb	Tunisian Arabic	Arab	Africa	اللغة العربيّة التونسيّة	Arabic (Tunisia); Tounsi	66	862	3998
en_af	Afrikaans	Latn	Africa	Afrikaans		130	863	0
en_ahr	Ahirani	Deva	Asia	खान्देशी, अहिराणी		457	0	0
en_aii	Assyrian Neo-Aramaic	Syrc	Asia	ܣܘܪܝܬ	Assyrian; Assyrian Neo–Aramaic; Sureth; Suret	0	0	3717
en_ak	Twi	Latn	Africa	Twi	Akan (includes Akuapem + Fante); Asante	260	863	4000
en_alz	Alur	Latn	Africa	Lur		66	863	3992
en_am	Amharic	Ethi	Africa	አማርኛ		584	863	0
en_apc	North Levantine Arabic	Arab	Asia	شامي‎		0	0	3998
en_apd	Sudanese Arabic	Arab	Africa	عربي سوداني	Arabic (Sudan)	0	0	0
en_ar-MA	Morrocan Arabic	Arab	Africa	الدارجة	Arabic (Morocco); Darija	260	863	3989
en_arn	Mapudungun	Latn	Americas	Mapudungun	Mapuche	0	0	4001
en_arz	Egyptian Arabic	Arab	Africa	اللهجة المصرية	Arabic (Egypt); Masri; Masry	260	863	4001
en_as	Assamese	Beng	Asia	অসমীয়া		0	0	3975
en_av	Avar	Cyrl	Europe	магӏарул мацӏ	Avaric	0	0	4001
en_awa	Awadhi	Deva	Asia	अवधी	Audhi	0	0	3999
en_ay	Aymara	Latn	Americas	Aymar aru		0	863	3990
en_ayl	Libyan Arabic	Arab	Africa	العربية الليبية	Arabic (Libya); Sulaimitian Arabic	66	863	3999
en_ba	Bashkir	Cyrl	Europe	Башҡорт		0	0	3986
en_bal	Baluchi	Arab	Asia	بلوچی	Balochi	0	0	3994
en_ban	Balinese	Latn	Asia	Basa Bali		0	0	3992
en_bbc	Batak Toba	Latn	Asia	Batak Toba		0	0	3993
en_bci	Baoulé	Latn	Africa	wawle	Baoule; Baule	66	863	3998
en_bem	Bemba (Zambia)	Latn	Africa	Ichibemba	Bemba; chiBemba	66	863	3977
en_ber	Tamazight (Tifinagh Script)	Tfng	Africa	ⵜⴰⵎⴰⵣⵉⵖⵜ	Berber; Amazigh	130	862	4000
en_ber-Latn	Tamazight (Latin Script)	Latn	Africa	Tamaziɣt	Tamazight; Berber; Amazigh	130	862	4001
en_bew	Betawi	Latn	Asia	Betawi	Betawi Malay; Jakartan Malay; Batavian Malay	0	0	4001
en_bfq	Badaga	Taml	Asia	படக		457	0	0
en_bfy	Bagheli	Deva	Asia	बघेली		457	0	0
en_bgq	Bagri	Deva	Asia	बागड़ी		457	0	0
en_bho	Bhojpuri	Deva	Asia	भोजपुरी		0	0	3991
en_bik	Bikol	Latn	Asia	Sentral Bikol	Bicol; Bicolano	0	0	3993
en_bjn	Banjar	Latn	Asia	Banjar	Banjarese	0	0	3996
en_bjn-Arab	Banjar (Arabic script)	Arab	Asia	بنجر; ڤندير بنجر		0	0	3999
en_bm	Bambara	Latn	Africa	Bámánánkán		260	863	4000
en_bns	Bundeli	Deva	Asia	बुन्देली		456	0	0
en_bo	Tibetan	Tibt	Asia	བོད་མི།; བོད་ཀྱི།; བོད་སྐད།; བོད་ཡིག།	Lhasa Tibetan; Standard Tibetan; Baykha; Kham; Baykay	0	863	3994
en_br	Breton	Latn	Europe	brezhoneg		0	0	4001
en_bra	Braj	Deva	Asia	ब्रजभाषा		457	0	0
en_brx	Bodo (India)	Deva	Asia	बोडो	Bodo; Boro	0	863	3994
en_bts	Batak Simalungun	Latn	Asia	Bahasa Simalungun	Simalungun; Simalungunese	0	0	3999
en_btx	Batak Karo	Latn	Asia	Batak Karo	Karonese; Karo; Bahasa Karo	0	0	3997
en_bua	Buryat	Cyrl	Asia	Буряад		0	0	3976
en_bug	Buginese	Latn	Asia	basa Ugi		0	0	3759
en_ccp-Latn	Chakma (Latin script)	Latn	Asia	Changhma Bhach		457	0	0
en_ce	Chechen	Cyrl	Europe	Нохчийн		0	0	3997
en_cgg	Kiga	Latn	Africa	Kiga	Chiga; Orukiga; Rukiga; Ruchiga	66	863	4001
en_ch	Chamorro	Latn	Oceania	Finuʼ Chamoru	CHamoru	0	0	4001
en_chk	Chuukese	Latn	Oceania	Trukese	Trukese	0	0	3889
en_chm	Meadow Mari	Cyrl	Asia	олык марий	Eastern Mari	0	0	4001
en_ckb	Kurdish (Sorani)	Arab	Asia	سۆرانی	Central Kurdish; Sorani	0	0	3986
en_cnh	Hakha Chin	Latn	Asia	Laica	Zomi; Laiholh; Lai	0	0	3999
en_crh	Crimean Tatar (Cyrillic script)	Cyrl	Europe	Крымскотатарский	Crimean Tatar	0	0	3990
en_crh-Latn	Crimean Tatar (Latin script)	Latn	Europe	qırımtatar tili		0	0	3990
en_crs	Seychellois Creole	Latn	Africa	kreol seselwa	kreol; Seselwa Creole French	0	0	3989
en_ctg	Chittagonian	Beng	Asia	Chittagonian		0	0	3996
en_cv	Chuvash	Cyrl	Europe	Чăваш		0	0	3985
en_dhd	Dhundari	Deva	Asia	ढूंढाड़ी		456	0	0
en_din	Dinka	Latn	Africa	Thuɔŋjäŋ		66	863	4001
en_doi	Dogri	Deva	Asia	डोगरी		454	0	3997
en_dov	Dombe	Latn	Africa	ChiDombe	Chidombe; Zambezi; Tonga	66	863	3998
en_dv	Dhivehi	Thaa	Asia	ދިވެހި	Maldivian, Divehi	0	0	3973
en_dyu	Dyula	Latn	Africa	Julakan	Dioula; Jula	66	863	3997
en_dz	Dzongkha	Tibt	Asia	རྫོང་ཁ	Bhutanese	0	0	3997
en_ee	Ewe	Latn	Africa	Eʋegbe		130	863	3990
en_efi	Efik	Latn	Africa	Usem mbom Efik; Ikɔ-efik		66	863	4001
en_es	Spanish	Latn	Europe	español, castellano		0	863	3991
en_fa-AF	Dari	Arab	Asia	دری	Persian (Afghanistan)	0	0	4000
en_ff	Fulani	Latn	Africa	Fulɓe; Fulfulde; Pular; Pulaar	Fulfulde; Fula; Fulah; Pular; Pulaar; Peul (French)	260	862	3998
en_fj	Fijian	Latn	Oceania	Fiji; Fijian	Bauan	0	0	4001
en_fo	Faroese	Latn	Europe	Føroysk		0	0	3999
en_fon	Fon	Latn	Africa	fongbē		66	863	3998
en_fr	French	Latn	Europe	français		0	0	3998
en_fr-CA	French (Canada)	Latn	Americas	français canadien		0	0	4001
en_fur	Friulian	Latn	Europe	Furlan		0	0	3985
en_gaa	Ga	Latn	Africa	Gã		66	863	4000
en_gn	Guarani	Latn	Americas	avañeʼẽ		0	863	3991
en_gom	Goan Konkani	Deva	Asia	कोंकणी		0	0	3999
en_grt-Latn	Garo (Latin script)	Latn	Asia	A·chikku		457	0	0
en_ha	Hausa	Latn	Africa	Hausa		584	863	0
en_hi	Hindi	Deva	Asia	हिन्दी		0	0	3984
en_hil	Hiligaynon	Latn	Asia	Hiligaynon	Ilonggo; Binisaya; Bisaya nga Hiniligaynon; Inilonggo	0	0	4001
en_hne	Chhattisgarhi	Deva	Asia	छत्तीसगढ़ी		0	0	4000
en_hoc-Wara	Ho (Warang Chiti script)	Wara	Asia	𑢹𑣉𑣉 𑣎𑣋𑣜		457	0	0
en_hrx	Hunsrik	Latn	Americas	Hunsrik	Riograndenser Hunsrik	0	0	3932
en_iba	Iban	Latn	Asia	Iban		0	0	3997
en_ig	Igbo	Latn	Africa	Asụsụ Igbo		391	863	0
en_ilo	Ilocano	Latn	Asia	Iloko	Iloko; Ilokano	0	0	3993
en_iso	Isoko	Latn	Africa	Isoko		0	0	4001
en_iu	Inuktut (Syllabics)	Cans	Americas	ᐃᓄᒃᑎᑐᑦ	Inuktitut; Inuktut	0	0	3999
en_iu-Latn	Inuktut (Latin)	Latn	Americas	Inuktitut	Inuktitut; Inuktut	0	0	4001
en_jam	Jamaican Patois	Latn	Americas	Patwa	Patois; Jamaican Creole English	0	0	3996
en_kaa	Karakalpak	Cyrl	Asia	Қарақалпақ; Қарақалпақша	Kara-Kalpak; Qaraqalpaq	0	0	3970
en_kac	Jingpo	Latn	Asia	Jinghpaw	Kachin; Jingpho; Jingphaw; Wunpawng	0	0	3941
en_kbd	East Circassian	Cyrl	Europe	Къэбэрдей; Адыгэ	Kabardian	0	0	3976
en_kek	Q'eqchi'	Latn	Americas	Qʼeqchiʼ	Kekchi; Kekchi'; Kekchí	0	0	4001
en_kfy	Kumaoni	Deva	Asia	कुमाऊँनी		457	0	0
en_kg	Kongo	Latn	Africa	Kikongo	Kikongo	66	863	3998
en_kha	Khasi	Latn	Asia	khasi		0	0	4001
en_ki	Kikuyu	Latn	Africa	Gĩgĩkũyũ	Gikuyu	66	863	3994
en_kl	Kalaallisut	Latn	Americas	Kalaallisut	Greenlandic	0	863	3943
en_kr	Kanuri	Latn	Africa	Kànùrí		66	863	4000
en_kri	Krio	Latn	Africa	Krio		130	863	4020
en_kru	Kurukh	Deva	Asia	कुँड़ुख़		457	0	0
en_ks	Kashmiri	Arab	Asia	كٲشُر	Koshur	0	863	4001
en_ks-Deva	Kashmiri (Devanagari script)	Deva	Asia	कॉशुर	Koshur	0	863	4001
en_ktu	Kituba (DRC)	Latn	Africa	Kituba	Kituba	66	863	3999
en_kv	Komi	Cyrl	Asia	Коми	Zyryan; Komi-Zyryan	0	0	3980
en_lep	Lepcha	Lepc	Asia	ᰛᰩᰵᰛᰧᰵᰶ‎		456	0	0
en_lg	Luganda	Latn	Africa	Oluganda	Ganda	260	863	3998
en_li	Limburgish	Latn	Europe	Limburgs	Limburgian; Limburg; Limburgs	0	0	3999
en_lif-Limb	Limbu (Limbu script)	Limb	Asia	ᤕᤠᤰᤌᤢᤱ ᤐᤠᤴ		457	0	0
en_lij	Ligurian	Latn	Europe	Ligure	Genoese	0	0	3976
en_lmo	Lombard	Latn	Europe	lombard		0	0	3983
en_ln	Lingala	Latn	Africa	Lingála	Ngala	260	863	3994
en_ltg	Latgalian	Latn	Europe	latgalīšu		0	0	3984
en_lu	Kiluba (Luba-Katanga)	Latn	Africa	Kiluba	Luba-Katanga; Luba; Tshiluba	66	863	0
en_lua	Tshiluba (Luba-Lulua)	Latn	Africa	Tshiluba, Ciluba	Luba-Kasai; Luba; Tshiluba; Cilubà; Tshilubà, Luba-Lulua	0	0	3998
en_luo	Luo	Latn	Africa	Luo	Dholuo	260	863	3941
en_lus	Mizo	Latn	Asia	Mizo ṭawng		0	863	3998
en_mad	Madurese	Latn	Asia	Madurâ		0	0	3999
en_mag	Magahi	Deva	Asia	मगही		456	0	3997
en_mai	Maithili	Deva	Asia	मैथिली		0	0	3989
en_mak	Makassar	Latn	Asia	Mangkasara	Makasar; Makassarese; Mangkasara; Macassar	0	0	3994
en_mam	Mam	Latn	Americas	Ta yol Mam		0	0	3902
en_meo	Kedah Malay	Latn	Asia	Siti	Malay (Kedah)	0	0	3995
en_mfe	Mauritian Creole	Latn	Africa	Morisien	Morisien; Morisyen	66	863	3997
en_mg	Malagasy	Latn	Africa	Fiteny malagasy		391	863	0
en_mh	Marshallese	Latn	Oceania	Majol	Ebon	0	0	3977
en_min	Minang	Latn	Asia	Minangkabau	Minangkabau	0	0	3999
en_mjl	Mandeali	Deva	Asia	मण्डियाली		457	0	0
en_mni-Mtei	Meiteilon (Manipuri)	Mtei	Asia	ꯃꯤꯇꯩꯂꯣꯟ	Meitei, Meetei, Meitheilon, Meeteilon	0	863	3923
en_mos	Mooré	Latn	Africa	Moaaga; Mòoré	Mossi; Moore; More; Mõore; Mossé	66	863	4000
en_ms-Arab	Malay (Jawi Script)	Arab	Asia	ملايو		0	0	4000
en_mtr	Mewari	Deva	Asia	मेवाड़ी		457	0	0
en_mwr	Marwadi	Deva	Asia	मारवाड़ी	Marwari; Rajasthani	0	0	3999
en_nd	North Ndebele	Latn	Africa	IsiNdebele	Ndebele (North)	66	863	3694
en_ndc	Ndau	Latn	Africa	Cindawu		66	863	3991
en_ne	Nepali	Deva	Asia	नेपाली		456	0	0
en_new	Nepalbhasa (Newari)	Deva	Asia	नेवाः	Newari; Nepalbhasa; Nepal Bhasa; Newar	0	0	4001
en_nhe	Eastern Huasteca Nahuatl	Latn	Americas	Nahuatlajtoli Kuextekan ikisayo	Nahuatl (Eastern Huasteca)	0	0	4000
en_noe	Nimadi	Deva	Asia	निमाड़ी		315	0	0
en_nqo	NKo	Nkoo	Africa	ߓߡߊߣߊ߲ߞߊ߲	N'Ko	0	0	0
en_nr	South Ndebele	Latn	Africa	isiNdebele seSewula	Ndebele (South); Transvaal Ndebele; Southern Ndebele	66	863	4001
en_nso	Sepedi	Latn	Africa	Sepedi	Northern Sotho; Pedi	130	863	3996
en_nus	Nuer	Latn	Africa	Thok Naath		0	0	4001
en_nv	Navajo	Latn	Americas	Diné; Naabeehó	Navaho; Diné	0	0	3992
en_ny	Chichewa	Latn	Africa	Chilankhulo cha Chichewa		260	863	0
en_oc	Occitan	Latn	Europe	Occitan	Provençal	0	0	3998
en_om	Oromo	Latn	Africa	Afaan Oromoo		391	863	3995
en_os	Ossetian	Cyrl	Europe	Ирон	Ossetic; Iron	0	0	3915
en_pa-Arab	Lahnda Punjabi (Pakistan)	Arab	Asia	لہندا پنجابی	Punjabi (Lahnda); Punjabi (Western); Punjabi (India); Punjabi (Shahmukhi)	0	863	4001
en_pag	Pangasinan	Latn	Asia	Pangasinan	Pangasinense	0	0	3994
en_pam	Kapampangan	Latn	Asia	Kapampangan	Pampango; Capampáñgan	0	0	3993
en_pap	Papiamento	Latn	Americas	Papiamento	Papiamento in Aruba; Papiamentu in Bonaire and Curaçao	0	0	4001
en_pcm	Nigerian Pidgin	Latn	Africa	Naijá, Pidgin for Nigerian	Naija; Naijá; Pijin; Brokun; Broken	130	863	4001
en_pt-PT	Portuguese (Portugal)	Latn	Europe	português		0	0	3995
en_qu	Quechua	Latn	Americas	Runa Simi		0	863	3861
en_quc	K'iche'	Latn	Americas	K'iche'	Quiche; Quiché	0	0	3992
en_rhg-Latn	Rohingya (Latin script)	Latn	Asia	Ruháingga	Rohingya	0	0	3999
en_rn	Rundi	Latn	Africa	Ikirundi	Kirundi	66	863	3999
en_rom	Romani	Latn	Europe	Rromani ćhib	Romany; Romanes; Roma	0	0	3988
en_rw	Kinyarwanda	Latn	Africa	Ikinyarwanda		260	863	0
en_sa	Sanskrit	Deva	Asia	संस्कृतम्		0	863	3998
en_sah	Yakut	Cyrl	Europe	Саха	Sakha; Saxa; Saqa; Yakutian	0	0	3983
en_sat	Santali (Ol Chiki script)	Olck	Asia	ᱥᱟᱱᱛᱟᱲᱤ	Santali; Santhali; Hor	0	863	4001
en_sat-Latn	Santali (Latin Script)	Latn	Asia	Santali	Santali; Santhali; Hor	0	863	4000
en_scl	Shina	Arab	Asia	ݜݨیاٗ زبان		457	0	0
en_scn	Sicilian	Latn	Europe	Sicilianu		100	0	3997
en_sd-Deva	Sindhi (Devanagari script)	Deva	Asia	सिन्धी		456	0	0
en_se	Northern Sami	Latn	Europe	davvisámegiella	North Saami; North Sámi; Sami (North)	0	0	3980
en_sg	Sango	Latn	Africa	Sango		0	0	3994
en_sgj	Surgujia	Deva	Asia	सरगुजिया		356	0	0
en_shn	Shan	Mymr	Asia	တႆး; လိၵ်ႈတႆး	Tai longe	0	0	4001
en_sjp	Surjapuri	Deva	Asia	सुरजापुरी‎		299	0	0
en_skr	Saraiki	Arab	Asia	سرائیکی	Multani	0	0	4001
en_sn	Shona	Latn	Africa	ChiShona		260	863	0
en_so	Somali	Latn	Africa	Af-Soomaali		260	862	0
en_spv	Sambalpuri	Orya	Asia	ସମ୍ବଲପୁରୀ		457	0	0
en_ss	Swati	Latn	Africa	Siswati	siSwati	66	863	3924
en_st	Sesotho	Latn	Africa	seSotho		260	863	0
en_sus	Susu	Latn	Africa	Sosokhui		66	863	4000
en_sw	Swahili	Latn	Africa	Kiswahili	Kiswahili	584	863	0
en_syl	Sylheti	Beng	Asia	সিলেটি		0	0	4001
en_szl	Silesian	Latn	Europe	Ślůnski		0	0	3997
en_tcy	Tulu	Knda	Asia	ತುಳು		451	0	4000
en_tet	Tetum	Latn	Asia	Tetun		0	0	4001
en_ti	Tigrinya	Ethi	Africa	ትግርኛ	Tigrigna	260	863	4000
en_tiv	Tiv	Latn	Africa	Tiv		66	863	4001
en_tn	Tswana	Latn	Africa	Setswana	Batswana; Setswana	66	863	3995
en_to	Tonga (Tonga Islands)	Latn	Oceania	Tonga	Tongan; Tonga; Lea Faka-Tonga	0	0	4001
en_tpi	Tok Pisin	Latn	Oceania	Tok Pisin	Papua New Guinea Pidgin	0	0	4001
en_trp	Kokborok	Latn	Asia	Kok Borok	Tripuri; Tiprakok; Kok Borok; Kókborok	457	0	3997
en_ts	Tsonga	Latn	Africa	Xitsonga		66	863	3994
en_tum	Tumbuka	Latn	Africa	Chitumbuka		66	863	3983
en_ty	Tahitian	Latn	Oceania	Reo Tahiti; Tahiti		0	0	4001
en_tyv	Tuvan	Cyrl	Asia	тувинский	Tuvinian	0	0	3984
en_udm	Udmurt	Cyrl	Europe	Удмурт кыл		0	0	3992
en_unr-Deva	Mundari (Devanagari script)	Deva	Asia	मुंडारी		457	0	0
en_ve	Venda	Latn	Africa	Tshivenḓa	Tshivenda; Luvenda; Bavenda	66	863	3984
en_vec	Venetian	Latn	Europe	Veneto		0	0	4000
en_war	Waray (Philippines)	Latn	Asia	Waray	Waray; Waray-Waray	0	0	3963
en_wbr	Wagdi	Deva	Asia	वाग्दी‎		455	0	0
en_wo	Wolof	Latn	Africa	Wolof		260	863	4001
en_xh	Xhosa	Latn	Africa	isiXhosa		260	863	0
en_xnr	Kangri	Deva	Asia	कांगड़ी		457	0	0
en_xsr-Tibt	Sherpa (Tibetan script)	Tibt	Asia	ཤར་པའི་སྐད་ཡིག		457	0	0
en_yo	Yoruba	Latn	Africa	Èdè Yorùbá		584	863	0
en_yua	Yucatec Maya	Latn	Americas	Maayat'aan	Mayan; Yucateco	0	0	3959
en_yue	Cantonese	Hant	Asia	廣東話	Yue Chinese	584	863	4201
en_zap	Zapotec	Latn	Americas	Diidxazá		0	0	3777
en_zu	Zulu	Latn	Africa	isiZulu		260	863	0
en_zza	Zazaki	Latn	Asia	Zazakî	Zaza; Kurdish (Zazaki); Dimli	0	0	3978
sw_ar	Arabic	Arab	Asia	اللُّغَة العَرَبِيّة	Arabic (MSA); Arabic (Standard); Arabic (Fusha)	330	0	0
sw_zh	Mandarin Chinese	Hans	Asia	汉语		330	0	0
yue_en	English	Latn	Europe	English		584	863	3983
yue_zh	Mandarin Chinese	Hans	Asia	汉语		0	0	4201
yue_zh-Hant	Taiwanese Mandarin	Hant	Asia	中文（台灣）		0	0	4201
zh-Hant_yue	Cantonese	Hant	Asia	廣東話	Yue Chinese	0	0	4031
zh_yue	Cantonese	Hant	Asia	廣東話	Yue Chinese	0	0	4031

Notes on language codes and names

We use the backwards-compatible BCP-47 language code convention. Several of these languages have various alternate code also in use, including:

• ak Twi also uses the code tw. Note that using the supercode for the most prevelent language variety is very common, e.g. ku for kmr, zh for cmn, etc.
• ber The grouping of the Tamazight languages is a matter of debate. The variety in these data is probably close to tzm, though a language expert should opine
• qu Another supercode issue. The Quechua variety this data is closest to is likely quy, though a language expert should opine
• pa-Arab is also referred to as lah
• chm: we use chm for Meadow Mari, although chm is properly the superset, and mhr is the individual code for Meadow Mari.
• rom: Similarly, we use rom for all Romani dialects, ak for Twi (see note above), and ff for all Fulfulde dialects.
• fa-AF,ar-MA: We use fa-AF and ar-MA for Dari and Darija respectively. The specific codes prs and ary are probably more precise.

Please note that the language name can be a point of considerable tension, and there is rarely a "right" name for a language! This said, here are some non-exhaustive notes that went into the names in the second column above, based on conversations with speakers. We welcome more comments from the community:

• ak: Calling this language "Twi" instead of "Akan" here is reflective of the corresponding translation models active on Google Translate. These are likely to mix different Akan dialects, including Twi proper, Fante, Akuapem Twi, etc.. Therefore, from a linguistic perspective, one would call this "Akan" and use the corresponding code "ak". However, this is a more academic word for the language; according to native speaker consultants, "Akan" is usually used to describe the ethnic group rather than the language, so they suggested simply calling this "Twi", which is most likely accurate for the large majority of model outputs.
• ckb: Central Kurdish is a more academic name that may not be as widely known to speakers of the language, so Sorani is preferred -- though they are both acceptable.
• dv: There is not a consistent romanization of this language, and both Divehi and Dhivehi are common. The first consonant in this language is a dental, non-aspirated d. When transliterating Indian language names there is no standard about how to represent this sound. In the south, "dh" usually refers to the dental consonant; in the North, it just as often refers to the aspirated one. However, in both cases, it is common that "d" alone would represent the retroflex consonant, so it is possible that Divehi would be mispronounced more often.
• gom: Technically, gom is the code for Goan Konkani only. Other varieties of Konkani, like Mangalore Konkani (which oddly enough does not have an ISO code), are mutually unintelligible! There is certainly dispute about whether to treat Konkani as one language or not.
• ilo: Native speakers give a variety of different answers around the correct name. Some say that Ilocano is more formal, whereas Ilokano is more informal; some say that both of these actually refer to the people, and that Iloko is the only correct term for the language.
• lg: Speakers usually refer to this language as "Luganda"; linguists sometimes call it "Ganda", but like Sepedi, this is not usually used by people familiar with the language.
• lus: There is agreement that Lushai is an obsolete colonial name and should not be used. Mizo is the preferred name.
• mni-Mtei: The conventional and most common name for this language is Manipuri. However, in reality, this is only the language of one of the many ethnicities in Manipur, so that name can be seen as offensive to some. Therefore, we use the endonym Meiteilon, with the conventional name (Manipuri) in parentheses.
• nso: Although a linguist might say that Sepedi is a dialect of Northern Sotho, speakers often say that they call the whole language Sepedi, and that Northern Sotho has offensive colonial overtones. Others, however, may disagree with this. The endonym Sesotho sa Leboa is not uncommon. There is agreement that no-one uses Pedi.
• qu: These translations are technically in Southern Quechua. However, this is a point of some contention. Some will opine that "Quechua" is not a language but a family of languages. Others, however, say that they all feel like one language, and that the regional differences are slight, and it's easy for Peruvians, Bolivians, and even Ecuadorians to understand each other. The majority opinion among the naive speakers we talked to seemed to be closer to the latter, so we have opted for the name "Quechua".
• ti: In Tigray Tigrigna is the most common Romanization, and in Eritrea Tigrinya is.
• zap: Linguists usually talk about the Zapotec languages, rather than considering Zapotec a single language. However, just as with Quechua above, conversations with a native speaker indicated that they perceived Zapotec as one language, with a variety of local variants that are all easy enough to understand (excepting one or two). A broader survey would help resolve this question.
• Those languages in this dataset that are not on Google Translate have yet to undergo a more careful name review; if you see a name in the list above that you want to correct, please contact us!

Special note on Bantu languages: Bantu language names have a special noun-class prefix for languages, like xiTsonga, sePedi, kiSwahili, isiZulu, luGanda, siSwati, isiNdebele, tshiVenda, and so on. However, there is no pattern as to whether this prefix is used in the conventional English form of the name. For instance, Xhosa usually leaves off the isi- in English, whereas Lingala almost never leaves off the li-. Others, like (xi)Tsonga, seem to be frequently used in both forms. Linguistic resources occasionally strip the prefixes, yielding names like "Ganda", "Pedi" and so on, perhaps out of a desire for consistent handling of the noun class prefix. (Even the ba in Bantu is such a prefix, so these languages are sometimes called the Ntu languages!)

For convenience, here is a mapping from codes used by Flores and SMOL:

flores2smol = {
'ace_Arab': 'ace-Arab', 'ace_Latn': 'ace', 'acm_Arab': 'acm', 'acq_Arab': 'acq', 'aeb_Arab': 'aeb', 'afr_Latn': 'af', 'ajp_Arab': 'ajp',
'aka_Latn': 'ak', 'als_Latn': 'sq', 'amh_Ethi': 'am', 'apc_Arab': 'apc', 'arb_Arab': 'ar', 'arb_Latn': 'ar-Latn', 'ars_Arab': 'ars',
'ary_Arab': 'ar-MA', 'arz_Arab': 'arz', 'asm_Beng': 'as', 'ast_Latn': 'ast', 'awa_Deva': 'awa', 'ayr_Latn': 'ay', 'azb_Arab': 'azb',
'azj_Latn': 'az', 'bak_Cyrl': 'ba', 'bam_Latn': 'bm', 'ban_Latn': 'ban', 'bel_Cyrl': 'be', 'bem_Latn': 'bem', 'ben_Beng': 'bn',
'bho_Deva': 'bho', 'bjn_Arab': 'bjn-Arab', 'bjn_Latn': 'bjn', 'bod_Tibt': 'bo', 'bos_Latn': 'bs', 'bug_Latn': 'bug', 'bul_Cyrl': 'bg',
'cat_Latn': 'ca', 'ceb_Latn': 'ceb', 'ces_Latn': 'cs', 'cjk_Latn': 'cjk', 'ckb_Arab': 'ckb', 'crh_Latn': 'crh-Latn', 'cym_Latn': 'cy',
'dan_Latn': 'da', 'deu_Latn': 'de', 'dik_Latn': 'din', 'dyu_Latn': 'dyu', 'dzo_Tibt': 'dz', 'ell_Grek': 'el', 'eng_Latn': 'en',
'epo_Latn': 'eo', 'est_Latn': 'et', 'eus_Latn': 'eu', 'ewe_Latn': 'ee', 'fao_Latn': 'fo', 'fij_Latn': 'fj', 'fin_Latn': 'fi',
'fon_Latn': 'fon', 'fra_Latn': 'fr', 'fur_Latn': 'fur', 'fuv_Latn': 'ff', 'gaz_Latn': 'om', 'gla_Latn': 'gd', 'gle_Latn': 'ga',
'glg_Latn': 'gl', 'grn_Latn': 'gn', 'guj_Gujr': 'gu', 'hat_Latn': 'ht', 'hau_Latn': 'ha', 'heb_Hebr': 'iw', 'hin_Deva': 'hi',
'hne_Deva': 'hne', 'hrv_Latn': 'hr', 'hun_Latn': 'hu', 'hye_Armn': 'hy', 'ibo_Latn': 'ig', 'ilo_Latn': 'ilo', 'ind_Latn': 'id',
'isl_Latn': 'is', 'ita_Latn': 'it', 'jav_Latn': 'jv', 'jpn_Jpan': 'ja', 'kab_Latn': 'kab', 'kac_Latn': 'kac', 'kam_Latn': 'kam',
'kan_Knda': 'kn', 'kas_Arab': 'ks', 'kas_Deva': 'ks-Deva', 'kat_Geor': 'ka', 'kaz_Cyrl': 'kk', 'kbp_Latn': 'kbp', 'kea_Latn': 'kea',
'khk_Cyrl': 'mn', 'khm_Khmr': 'km', 'kik_Latn': 'ki', 'kin_Latn': 'rw', 'kir_Cyrl': 'ky', 'kmb_Latn': 'kmb', 'kmr_Latn': 'ku',
'knc_Arab': 'kr-Arab', 'knc_Latn': 'kr', 'kon_Latn': 'kg', 'kor_Hang': 'ko', 'lao_Laoo': 'lo', 'lij_Latn': 'lij', 'lim_Latn': 'li',
'lin_Latn': 'ln', 'lit_Latn': 'lt', 'lmo_Latn': 'lmo', 'ltg_Latn': 'ltg', 'ltz_Latn': 'lb', 'lua_Latn': 'lua', 'lug_Latn': 'lg',
'luo_Latn': 'luo', 'lus_Latn': 'lus', 'lvs_Latn': 'lv', 'mag_Deva': 'mag', 'mai_Deva': 'mai', 'mal_Mlym': 'ml', 'mar_Deva': 'mr',
'min_Arab': 'min-Arab', 'min_Latn': 'min', 'mkd_Cyrl': 'mk', 'mlt_Latn': 'mt', 'mni_Beng': 'mni', 'mos_Latn': 'mos', 'mri_Latn': 'mi',
'mya_Mymr': 'my', 'nld_Latn': 'nl', 'nno_Latn': 'nn', 'nob_Latn': 'no', 'npi_Deva': 'ne', 'nso_Latn': 'nso', 'nus_Latn': 'nus',
'nya_Latn': 'ny', 'oci_Latn': 'oc', 'ory_Orya': 'or', 'pag_Latn': 'pag', 'pan_Guru': 'pa', 'pap_Latn': 'pap', 'pbt_Arab': 'ps',
'pes_Arab': 'fa', 'plt_Latn': 'mg', 'pol_Latn': 'pl', 'por_Latn': 'pt', 'prs_Arab': 'fa-AF', 'quy_Latn': 'quy', 'ron_Latn': 'ro',
'run_Latn': 'rn', 'rus_Cyrl': 'ru', 'sag_Latn': 'sg', 'san_Deva': 'sa', 'sat_Olck': 'sat', 'scn_Latn': 'scn', 'shn_Mymr': 'shn',
'sin_Sinh': 'si', 'slk_Latn': 'sk', 'slv_Latn': 'sl', 'smo_Latn': 'sm', 'sna_Latn': 'sn', 'snd_Arab': 'sd', 'som_Latn': 'so',
'sot_Latn': 'st', 'spa_Latn': 'es', 'srd_Latn': 'sc', 'srp_Cyrl': 'sr', 'ssw_Latn': 'ss', 'sun_Latn': 'su', 'swe_Latn': 'sv',
'swh_Latn': 'sw', 'szl_Latn': 'szl', 'tam_Taml': 'ta', 'taq_Latn': 'taq', 'taq_Tfng': 'taq-Tfng', 'tat_Cyrl': 'tt', 'tel_Telu': 'te',
'tgk_Cyrl': 'tg', 'tgl_Latn': 'fil', 'tha_Thai': 'th', 'tir_Ethi': 'ti', 'tpi_Latn': 'tpi', 'tsn_Latn': 'tn', 'tso_Latn': 'ts',
'tuk_Latn': 'tk', 'tum_Latn': 'tum', 'tur_Latn': 'tr', 'twi_Latn': 'ak', 'tzm_Tfng': 'ber', 'uig_Arab': 'ug', 'ukr_Cyrl': 'uk',
'umb_Latn': 'umb', 'urd_Arab': 'ur', 'uzn_Latn': 'uz', 'vec_Latn': 'vec', 'vie_Latn': 'vi', 'war_Latn': 'war', 'wol_Latn': 'wo',
'xho_Latn': 'xh', 'ydd_Hebr': 'yi', 'yor_Latn': 'yo', 'yue_Hant': 'yue', 'zho_Hans': 'zh', 'zho_Hant': 'zh-Hant', 'zsm_Latn': 'ms', 'zul_Latn': 'zu',}

Known issues

There are several known issues of this dataset:

• The translations of SmolDoc to Ho (Warang Citi script) (hoc) contain some instances of transliterated English, rather than actual Ho text. This is currently being corrected.
• The translations to Mooré (mos) contain a mix of orthographies.
• The translations into Kituba (ktu), Tulu (tcy), Nepali (ne), and Tibetan (bo) caused notable regressions in ChrF, suggesting that they are low-quality, though this could also be an issue with training or evaluation.
• The Manx (gv) split of GATITOS that was originally released was spurious. It has been removed.

Citations

To cite SMOL and GATITOS, please use the following:

@misc{caswell2025smol,
      title={SMOL: Professionally translated parallel data for 115 under-represented languages}, 
      author={Isaac Caswell and Elizabeth Nielsen and Jiaming Luo and Colin Cherry and Geza Kovacs and Hadar Shemtov and Partha Talukdar and Dinesh Tewari and Baba Mamadi Diane and Koulako Moussa Doumbouya and Djibrila Diane and Solo Farabado Cissé},
      year={2025},
      eprint={2502.12301},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2502.12301}, 
}

@misc{jones2023gatitos,
      title={Bilex Rx: Lexical Data Augmentation for Massively Multilingual Machine Translation}, 
      author={Alex Jones and Isaac Caswell and Ishank Saxena and Orhan Firat},
      year={2023},
      eprint={2303.15265},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2303.15265}, 
}

Contact

We welcome contributions, corrections, and suggestions, especially from speakers of these languages. If you speak any of these languages and notice quality issues, please file bugs on this repo or contact {icaswell, eknielsen}꩜google.com. This repository may be updated sporadically with corrections or additional translations.

License

This data is released under cc-by-4.0