Dataset Card for MAFAND

Dataset Summary

MAFAND-MT is the largest MT benchmark for African languages in the news domain, covering 21 languages.

Supported Tasks and Leaderboards

Machine Translation

Languages

The languages covered are:

• Amharic
• Bambara
• Ghomala
• Ewe
• Fon
• Hausa
• Igbo
• Kinyarwanda
• Luganda
• Luo
• Mossi
• Nigerian-Pidgin
• Chichewa
• Shona
• Swahili
• Setswana
• Twi
• Wolof
• Xhosa
• Yoruba
• Zulu

Dataset Structure

Data Instances

>>> from datasets import load_dataset
>>> data = load_dataset('masakhane/mafand', 'en-yor')

{"translation": {"src": "President Buhari will determine when to lift lockdown – Minister", "tgt": "Ààrẹ Buhari ló lè yóhùn padà lórí ètò kónílégbélé – Mínísítà"}}


{"translation": {"en": "President Buhari will determine when to lift lockdown – Minister", "yo": "Ààrẹ Buhari ló lè yóhùn padà lórí ètò kónílégbélé – Mínísítà"}}

Data Fields

• "translation": name of the task
• "src" : source language e.g en
• "tgt": target language e.g yo

Data Splits

Train/dev/test split

language	Train	Dev	Test
amh	-	899	1037
bam	3302	1484	1600
bbj	2232	1133	1430
ewe	2026	1414	1563
fon	2637	1227	1579
hau	5865	1300	1500
ibo	6998	1500	1500
kin	-	460	1006
lug	4075	1500	1500
luo	4262	1500	1500
mos	2287	1478	1574
nya	-	483	1004
pcm	4790	1484	1574
sna	-	556	1005
swa	30782	1791	1835
tsn	2100	1340	1835
twi	3337	1284	1500
wol	3360	1506	1500
xho	-	486	1002
yor	6644	1544	1558
zul	3500	1239	998

Dataset Creation

Curation Rationale

MAFAND was created from the news domain, translated from English or French to an African language

Source Data

Initial Data Collection and Normalization

[Needs More Information]

Who are the source language producers?

• Masakhane
• Igbo
• Swahili
• Hausa
• Yoruba

Annotations

Annotation process

[Needs More Information]

Who are the annotators?

Masakhane members

Personal and Sensitive Information

[Needs More Information]

Considerations for Using the Data

Social Impact of Dataset

[Needs More Information]

Discussion of Biases

[Needs More Information]

Other Known Limitations

[Needs More Information]

Additional Information

Dataset Curators

[Needs More Information]

Licensing Information

CC-BY-4.0-NC

Citation Information

@inproceedings{adelani-etal-2022-thousand,
    title = "A Few Thousand Translations Go a Long Way! Leveraging Pre-trained Models for {A}frican News Translation",
    author = "Adelani, David  and
      Alabi, Jesujoba  and
      Fan, Angela  and
      Kreutzer, Julia  and
      Shen, Xiaoyu  and
      Reid, Machel  and
      Ruiter, Dana  and
      Klakow, Dietrich  and
      Nabende, Peter  and
      Chang, Ernie  and
      Gwadabe, Tajuddeen  and
      Sackey, Freshia  and
      Dossou, Bonaventure F. P.  and
      Emezue, Chris  and
      Leong, Colin  and
      Beukman, Michael  and
      Muhammad, Shamsuddeen  and
      Jarso, Guyo  and
      Yousuf, Oreen  and
      Niyongabo Rubungo, Andre  and
      Hacheme, Gilles  and
      Wairagala, Eric Peter  and
      Nasir, Muhammad Umair  and
      Ajibade, Benjamin  and
      Ajayi, Tunde  and
      Gitau, Yvonne  and
      Abbott, Jade  and
      Ahmed, Mohamed  and
      Ochieng, Millicent  and
      Aremu, Anuoluwapo  and
      Ogayo, Perez  and
      Mukiibi, Jonathan  and
      Ouoba Kabore, Fatoumata  and
      Kalipe, Godson  and
      Mbaye, Derguene  and
      Tapo, Allahsera Auguste  and
      Memdjokam Koagne, Victoire  and
      Munkoh-Buabeng, Edwin  and
      Wagner, Valencia  and
      Abdulmumin, Idris  and
      Awokoya, Ayodele  and
      Buzaaba, Happy  and
      Sibanda, Blessing  and
      Bukula, Andiswa  and
      Manthalu, Sam",
    booktitle = "Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies",
    month = jul,
    year = "2022",
    address = "Seattle, United States",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2022.naacl-main.223",
    doi = "10.18653/v1/2022.naacl-main.223",
    pages = "3053--3070",
    abstract = "Recent advances in the pre-training for language models leverage large-scale datasets to create multilingual models. However, low-resource languages are mostly left out in these datasets. This is primarily because many widely spoken languages that are not well represented on the web and therefore excluded from the large-scale crawls for datasets. Furthermore, downstream users of these models are restricted to the selection of languages originally chosen for pre-training. This work investigates how to optimally leverage existing pre-trained models to create low-resource translation systems for 16 African languages. We focus on two questions: 1) How can pre-trained models be used for languages not included in the initial pretraining? and 2) How can the resulting translation models effectively transfer to new domains? To answer these questions, we create a novel African news corpus covering 16 languages, of which eight languages are not part of any existing evaluation dataset. We demonstrate that the most effective strategy for transferring both additional languages and additional domains is to leverage small quantities of high-quality translation data to fine-tune large pre-trained models.",
}