--- license: cc-by-4.0 task_categories: - translation pretty_name: Smol size_categories: - 10KxiTsonga, 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 and Edoardo Ferrante and Alessandro Guasoni and Mamadou K. Keita and Sudhamoy DebBarma and Ali Kuzhuget and David Anugraha and Muhammad Ravi Shulthan Habibi and Sina Ahmadi and Mingfei Lau and Jonathan Eng}, year={2025}, eprint={2502.12301}, archivePrefix={arXiv}, primaryClass={cs.CL}, url={https://arxiv.org/abs/2502.12301}, } @inproceedings{jones-etal-2023-gatitos, title = {{"GATITOS: Using a New Multilingual Lexicon for Low-resource Machine Translation"}}, author = "Jones, Alexander and Caswell, Isaac and Firat, Orhan and Saxena, Ishank", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.26/", doi = "10.18653/v1/2023.emnlp-main.26", pages = "371--405", abstract = "Modern machine translation models and language models are able to translate without having been trained on parallel data, greatly expanding the set of languages that they can serve. However, these models still struggle in a variety of predictable ways, a problem that cannot be overcome without at least some trusted bilingual data. This work expands on a cheap and abundant resource to combat this problem: bilingual lexica. We test the efficacy of bilingual lexica in a real-world set-up, on 200-language translation models trained on web-crawled text. We present several findings: (1) using lexical data augmentation, we demonstrate sizable performance gains for unsupervised translation; (2) we compare several families of data augmentation, demonstrating that they yield similar improvements, and can be combined for even greater improvements; (3) we demonstrate the importance of carefully curated lexica over larger, noisier ones, especially with larger models; and (4) we compare the efficacy of multilingual lexicon data versus human-translated parallel data. Based on results from (3), we develop and open-source GATITOS, a high-quality, curated dataset in 168 tail languages, one of the first human-translated resources to cover many of these languages." } ``` ### 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