Assessing back-translation as a corpus generation strategy for non-English tasks: A study in reading comprehension and word sense disambiguation

Fabricio Monsalve; Kervy Rivas-Rojas; Marco Antonio Sobrevilla Cabezudo; Arturo Oncevay-Marcos

Assessing back-translation as a corpus generation strategy for non-English tasks: A study in reading comprehension and word sense disambiguation

Fabricio Monsalve
, Kervy Rivas-Rojas
, Marco Antonio Sobrevilla Cabezudo
, Arturo Oncevay-Marcos

Pontificia Universidad Católica del Perú

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

1 Scopus citations

Abstract

Corpora curated by experts have sustained Natural Language Processing mainly in English, but the expensiveness of corpora creation is a barrier for the development in further languages. Thus, we propose a corpus generation strategy that only requires a machine translation system between English and the target language in both directions, where we filter the best translations by computing automatic translation metrics and the task performance score. By studying Reading Comprehension in Spanish and Word Sense Disambiguation in Portuguese, we identified that a more quality-oriented metric has high potential in the corpora selection without degrading the task performance. We conclude that it is possible to systematise the building of quality corpora using machine translation and automatic metrics, besides some prior effort to clean and process the data.

Original language	Spanish
Title of host publication	LAW 2019 - 13th Linguistic Annotation Workshop, Proceedings of the Workshop
Pages	81-89
Number of pages	9
State	Published - 1 Jan 2019

Cite this

@inproceedings{ce58fba13be643c7ad23e17ecc297a69,

title = "Assessing back-translation as a corpus generation strategy for non-English tasks: A study in reading comprehension and word sense disambiguation",

abstract = "Corpora curated by experts have sustained Natural Language Processing mainly in English, but the expensiveness of corpora creation is a barrier for the development in further languages. Thus, we propose a corpus generation strategy that only requires a machine translation system between English and the target language in both directions, where we filter the best translations by computing automatic translation metrics and the task performance score. By studying Reading Comprehension in Spanish and Word Sense Disambiguation in Portuguese, we identified that a more quality-oriented metric has high potential in the corpora selection without degrading the task performance. We conclude that it is possible to systematise the building of quality corpora using machine translation and automatic metrics, besides some prior effort to clean and process the data.",

author = "Fabricio Monsalve and Kervy Rivas-Rojas and Cabezudo, \{Marco Antonio Sobrevilla\} and Arturo Oncevay-Marcos",

year = "2019",

month = jan,

day = "1",

language = "Espa{\~n}ol",

pages = "81--89",

booktitle = "LAW 2019 - 13th Linguistic Annotation Workshop, Proceedings of the Workshop",

}

Assessing back-translation as a corpus generation strategy for non-English tasks: A study in reading comprehension and word sense disambiguation. / Monsalve, Fabricio; Rivas-Rojas, Kervy; Cabezudo, Marco Antonio Sobrevilla et al.
LAW 2019 - 13th Linguistic Annotation Workshop, Proceedings of the Workshop. 2019. p. 81-89.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Assessing back-translation as a corpus generation strategy for non-English tasks: A study in reading comprehension and word sense disambiguation

AU - Monsalve, Fabricio

AU - Rivas-Rojas, Kervy

AU - Cabezudo, Marco Antonio Sobrevilla

AU - Oncevay-Marcos, Arturo

PY - 2019/1/1

Y1 - 2019/1/1

N2 - Corpora curated by experts have sustained Natural Language Processing mainly in English, but the expensiveness of corpora creation is a barrier for the development in further languages. Thus, we propose a corpus generation strategy that only requires a machine translation system between English and the target language in both directions, where we filter the best translations by computing automatic translation metrics and the task performance score. By studying Reading Comprehension in Spanish and Word Sense Disambiguation in Portuguese, we identified that a more quality-oriented metric has high potential in the corpora selection without degrading the task performance. We conclude that it is possible to systematise the building of quality corpora using machine translation and automatic metrics, besides some prior effort to clean and process the data.

AB - Corpora curated by experts have sustained Natural Language Processing mainly in English, but the expensiveness of corpora creation is a barrier for the development in further languages. Thus, we propose a corpus generation strategy that only requires a machine translation system between English and the target language in both directions, where we filter the best translations by computing automatic translation metrics and the task performance score. By studying Reading Comprehension in Spanish and Word Sense Disambiguation in Portuguese, we identified that a more quality-oriented metric has high potential in the corpora selection without degrading the task performance. We conclude that it is possible to systematise the building of quality corpora using machine translation and automatic metrics, besides some prior effort to clean and process the data.

M3 - Contribución a la conferencia

SP - 81

EP - 89

BT - LAW 2019 - 13th Linguistic Annotation Workshop, Proceedings of the Workshop

ER -