Search: onr:"swepub:oai:DiVA.org:uu-489248" >
A Few Thousand Tran...
A Few Thousand Translations Go A Long Way! Leveraging Pre-trained Models for African News Translation
-
- Adelani, David (author)
- Saarland Univ, Saarbrucken, Germany.
-
- Alabi, Jesujoba (author)
- INRIA, Paris, France.
-
- Fan, Angela (author)
- Meta AI, Menlo Pk, CA USA.
-
show more...
-
- Kreutzer, Julia (author)
- Google Res, Mountain View, CA USA.
-
- Shen, Xiaoyu (author)
- Amazon Alexa AI, Seattle, WA USA.
-
- Reid, Machel (author)
- Univ Tokyo, Tokyo, Japan.
-
- Ruiter, Dana (author)
- Saarland Univ, Saarbrucken, Germany.
-
- Klakow, Dietrich (author)
- Saarland Univ, Saarbrucken, Germany.
-
- Nabende, Peter (author)
- Makerere Univ, Kampala, Uganda.
-
- Chang, Ernie (author)
- Saarland Univ, Saarbrucken, Germany.
-
- Gwadabe, Tajuddeen (author)
- UCAS, Beijing, Peoples R China.
-
- Sackey, Freshia (author)
- JKUAT, Juja, Kenya.
-
- Dossou, Bonaventure F. P. (author)
- Jacobs Univ, Bremen, Germany.
-
- Emezue, Chris (author)
- TUM, Munich, Germany.
-
- Leong, Colin (author)
- Univ Dayton, Dayton, OH 45469 USA.
-
- Beukman, Michael (author)
- Univ Witwatersrand, Johannesburg, South Africa.
-
- Muhammad, Shamsuddeen (author)
- LIAAD INESC TEC, Porto, Portugal.
-
Jarso, Guyo (author)
-
- Yousuf, Oreen (author)
- Uppsala universitet,Institutionen för lingvistik och filologi
-
- Rubungo, Andre Niyongabo (author)
- UPC, Barcelona, Spain.
-
- Hacheme, Gilles (author)
- Ai4Innov, Paris, France.
-
- Wairagala, Eric Peter (author)
- Makerere Univ, Kampala, Uganda.
-
- Nasir, Muhammad Umair (author)
- Ominor AI, Orlando, FL USA.
-
Ajibade, Benjamin (author)
-
Ajayi, Tunde (author)
-
Gitau, Yvonne (author)
-
Abbott, Jade (author)
-
- Ahmed, Mohamed (author)
- Microsoft Africa Res Inst, Nairobi, Kenya.
-
- Ochieng, Millicent (author)
- Microsoft Africa Res Inst, Nairobi, Kenya.
-
Aremu, Anuoluwapo (author)
-
- Ogayo, Perez (author)
- CMU, Pittsburgh, PA USA.
-
- Mukiibi, Jonathan (author)
- Makerere Univ, Kampala, Uganda.
-
Kabore, Fatoumata Ouoba (author)
-
Kalipe, Godson (author)
-
- Mbaye, Derguene (author)
- Baamtu, Dakar, Senegal.
-
- Tapo, Allahsera Auguste (author)
- RIT, Rochester, NY USA.
-
Koagne, Victoire Memdjokam (author)
-
Munkoh-Buabeng, Edwin (author)
-
- Wagner, Valencia (author)
- SPU, Kimberley, South Africa.
-
- Abdulmumin, Idris (author)
- ABU, Abuja, Nigeria.
-
- Awokoya, Ayodele (author)
- UI Ibadan, Ibadan, Nigeria.
-
Buzaaba, Happy (author)
-
- Sibanda, Blessing (author)
- NUST, Windhoek, Namibia.
-
- Bukula, Andiswa (author)
- SADiLaR, Potchefstroom, South Africa.
-
- Manthalu, Sam (author)
- Univ Malawi, Zomba, Malawi.
-
show less...
-
Saarland Univ, Saarbrucken, Germany INRIA, Paris, France. (creator_code:org_t)
- Stroudsburg : Association for Computational Linguistics, 2022
- 2022
- English.
-
In: NAACL 2022. - Stroudsburg : Association for Computational Linguistics. - 9781955917711 ; , s. 3053-3070
- Related links:
-
https://aclanthology...
-
show more...
-
https://urn.kb.se/re...
-
show less...
Abstract
Subject headings
Close
- Recent advances in the pre-training of 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 are not well represented on the web and therefore excluded from the large-scale crawls used to create 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 pre-training? and 2) How can the resulting translation models effectively transfer to new domains? To answer these questions, we create a new 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 to additional languages and to additional domains is to fine-tune large pre-trained models on small quantities of high-quality translation data.
Subject headings
- NATURVETENSKAP -- Data- och informationsvetenskap -- Språkteknologi (hsv//swe)
- NATURAL SCIENCES -- Computer and Information Sciences -- Language Technology (hsv//eng)
Publication and Content Type
- ref (subject category)
- kon (subject category)
Find in a library
-
NAACL 2022
(Search for host publication in LIBRIS)
To the university's database
- By the author/editor
-
Adelani, David
-
Alabi, Jesujoba
-
Fan, Angela
-
Kreutzer, Julia
-
Shen, Xiaoyu
-
Reid, Machel
-
show more...
-
Ruiter, Dana
-
Klakow, Dietrich
-
Nabende, Peter
-
Chang, Ernie
-
Gwadabe, Tajudde ...
-
Sackey, Freshia
-
Dossou, Bonavent ...
-
Emezue, Chris
-
Leong, Colin
-
Beukman, Michael
-
Muhammad, Shamsu ...
-
Jarso, Guyo
-
Yousuf, Oreen
-
Rubungo, Andre N ...
-
Hacheme, Gilles
-
Wairagala, Eric ...
-
Nasir, Muhammad ...
-
Ajibade, Benjami ...
-
Ajayi, Tunde
-
Gitau, Yvonne
-
Abbott, Jade
-
Ahmed, Mohamed
-
Ochieng, Millice ...
-
Aremu, Anuoluwap ...
-
Ogayo, Perez
-
Mukiibi, Jonatha ...
-
Kabore, Fatoumat ...
-
Kalipe, Godson
-
Mbaye, Derguene
-
Tapo, Allahsera ...
-
Koagne, Victoire ...
-
Munkoh-Buabeng, ...
-
Wagner, Valencia
-
Abdulmumin, Idri ...
-
Awokoya, Ayodele
-
Buzaaba, Happy
-
Sibanda, Blessin ...
-
Bukula, Andiswa
-
Manthalu, Sam
-
show less...
- About the subject
-
- NATURAL SCIENCES
-
NATURAL SCIENCES
-
and Computer and Inf ...
-
and Language Technol ...
- Articles in the publication
-
NAACL 2022
- By the university
-
Uppsala University