Help from the Neighbors: Estonian Dialect Normalization Using a Finnish Dialect Generator

Tuuli Tuisk; Hämäläinen Mika; Alnajjar Khalid

doi:10.18653/v1/2022.deeplo-1.7

Help from the Neighbors: Estonian Dialect Normalization Using a Finnish Dialect Generator

Tuuli Tuisk
, Hämäläinen Mika
, Alnajjar Khalid

Latvijas Universitāte

Ne LU

Zinātniskās darbības rezultāts: Nodaļa grāmatā/enciklopēdijā/konferences krājumā › Konferences zinātniskais raksts › Pētniecība › koleģiāli recenzēts

3 Atsauces (Scopus)

Kopsavilkums

While standard Estonian is not a low-resourced language, the different dialects of the language are under-resourced from the point of view of NLP, given that there are no vast hand normalized resources available for training a machine learning model to normalize dialectal Estonian to standard Estonian. In this paper, we crawl a small corpus of parallel dialectal Estonian - standard Estonian sentences. In addition, we take a savvy approach of generating more synthetic training data for the normalization task by using an existing dialect generator model built for Finnish to "dialectalize" standard Estonian sentences from the Universal Dependencies tree banks. Our BERT based normalization model achieves a word error rate that is 26.49 points lower when using both the synthetic data and Estonian data in comparison to training the model with only the available Estonian data. Our results suggest that synthetic data generated by a model trained on a more resourced related language can indeed boost the results for a less resourced language.

Oriģinālvaloda	Angļu
Rīkotāja publikācijas nosaukums	DeepLo 2022 - 3rd Workshop on Deep Learning Approaches for Low-Resource NLP, Proceedings of the DeepLo Workshop
Redaktori	Colin Cherry, Angela Fan, George Foster, Gholamreza Haffari, Shahram Khadivi, Nanyun Peng, Xiang Ren, Ehsan Shareghi, Swabha Swayamdipta
Publikācijas vieta	Seattle
Izdevējs	Association for Computational Linguistics
Lapas	61-66
Lapu skaits	6
ISBN (Elektroniski)	9781955917971
DOIs	https://doi.org/10.18653/v1/2022.deeplo-1.7
Publikācijas statuss	Publicēts - 2022

Publikāciju sērijas

Nosaukums	DeepLo 2022 - 3rd Workshop on Deep Learning Approaches for Low-Resource NLP, Proceedings of the DeepLo Workshop

OECD Zinātnes nozare

6.2 Valodniecība un literatūrzinātne

Piekļuve Dokumentiem

10.18653/v1/2022.deeplo-1.7

Citi faili un saites

Citēt šo

Tuisk, T., Mika, H., & Khalid, A. (2022). Help from the Neighbors: Estonian Dialect Normalization Using a Finnish Dialect Generator. (gadā) C. Cherry, A. Fan, G. Foster, G. Haffari, S. Khadivi, N. Peng, X. Ren, E. Shareghi, & S. Swayamdipta (Red.), DeepLo 2022 - 3rd Workshop on Deep Learning Approaches for Low-Resource NLP, Proceedings of the DeepLo Workshop (lpp. 61-66). (DeepLo 2022 - 3rd Workshop on Deep Learning Approaches for Low-Resource NLP, Proceedings of the DeepLo Workshop). Association for Computational Linguistics. https://doi.org/10.18653/v1/2022.deeplo-1.7

Tuisk, Tuuli ; Mika, Hämäläinen ; Khalid, Alnajjar. / Help from the Neighbors: Estonian Dialect Normalization Using a Finnish Dialect Generator. DeepLo 2022 - 3rd Workshop on Deep Learning Approaches for Low-Resource NLP, Proceedings of the DeepLo Workshop. redaktors / Colin Cherry ; Angela Fan ; George Foster ; Gholamreza Haffari ; Shahram Khadivi ; Nanyun Peng ; Xiang Ren ; Ehsan Shareghi ; Swabha Swayamdipta. Seattle : Association for Computational Linguistics, 2022. lpp. 61-66 (DeepLo 2022 - 3rd Workshop on Deep Learning Approaches for Low-Resource NLP, Proceedings of the DeepLo Workshop).

@inproceedings{d0798c1be5ad4a329a0f4cb9c9037599,

title = "Help from the Neighbors: Estonian Dialect Normalization Using a Finnish Dialect Generator",

abstract = "While standard Estonian is not a low-resourced language, the different dialects of the language are under-resourced from the point of view of NLP, given that there are no vast hand normalized resources available for training a machine learning model to normalize dialectal Estonian to standard Estonian. In this paper, we crawl a small corpus of parallel dialectal Estonian - standard Estonian sentences. In addition, we take a savvy approach of generating more synthetic training data for the normalization task by using an existing dialect generator model built for Finnish to {"}dialectalize{"} standard Estonian sentences from the Universal Dependencies tree banks. Our BERT based normalization model achieves a word error rate that is 26.49 points lower when using both the synthetic data and Estonian data in comparison to training the model with only the available Estonian data. Our results suggest that synthetic data generated by a model trained on a more resourced related language can indeed boost the results for a less resourced language.",

author = "Tuuli Tuisk and H{\"a}m{\"a}l{\"a}inen Mika and Alnajjar Khalid",

note = "Publisher Copyright: {\textcopyright} 2022 Association for Computational Linguistics.",

year = "2022",

doi = "10.18653/v1/2022.deeplo-1.7",

language = "English",

series = "DeepLo 2022 - 3rd Workshop on Deep Learning Approaches for Low-Resource NLP, Proceedings of the DeepLo Workshop",

publisher = "Association for Computational Linguistics",

pages = "61--66",

editor = "Colin Cherry and Angela Fan and George Foster and Gholamreza Haffari and Shahram Khadivi and Nanyun Peng and Xiang Ren and Ehsan Shareghi and Swabha Swayamdipta",

booktitle = "DeepLo 2022 - 3rd Workshop on Deep Learning Approaches for Low-Resource NLP, Proceedings of the DeepLo Workshop",

address = "United States",

}

Tuisk, T, Mika, H & Khalid, A 2022, Help from the Neighbors: Estonian Dialect Normalization Using a Finnish Dialect Generator. (kur) C Cherry, A Fan, G Foster, G Haffari, S Khadivi, N Peng, X Ren, E Shareghi & S Swayamdipta (red.), DeepLo 2022 - 3rd Workshop on Deep Learning Approaches for Low-Resource NLP, Proceedings of the DeepLo Workshop. DeepLo 2022 - 3rd Workshop on Deep Learning Approaches for Low-Resource NLP, Proceedings of the DeepLo Workshop, Association for Computational Linguistics, Seattle, lpp. 61-66. https://doi.org/10.18653/v1/2022.deeplo-1.7

Help from the Neighbors: Estonian Dialect Normalization Using a Finnish Dialect Generator. / Tuisk, Tuuli; Mika, Hämäläinen; Khalid, Alnajjar.
DeepLo 2022 - 3rd Workshop on Deep Learning Approaches for Low-Resource NLP, Proceedings of the DeepLo Workshop. red. / Colin Cherry; Angela Fan; George Foster; Gholamreza Haffari; Shahram Khadivi; Nanyun Peng; Xiang Ren; Ehsan Shareghi; Swabha Swayamdipta. Seattle: Association for Computational Linguistics, 2022. lpp. 61-66 (DeepLo 2022 - 3rd Workshop on Deep Learning Approaches for Low-Resource NLP, Proceedings of the DeepLo Workshop).

Zinātniskās darbības rezultāts: Nodaļa grāmatā/enciklopēdijā/konferences krājumā › Konferences zinātniskais raksts › Pētniecība › koleģiāli recenzēts

TY - GEN

T1 - Help from the Neighbors: Estonian Dialect Normalization Using a Finnish Dialect Generator

AU - Tuisk, Tuuli

AU - Mika, Hämäläinen

AU - Khalid, Alnajjar

PY - 2022

Y1 - 2022

N2 - While standard Estonian is not a low-resourced language, the different dialects of the language are under-resourced from the point of view of NLP, given that there are no vast hand normalized resources available for training a machine learning model to normalize dialectal Estonian to standard Estonian. In this paper, we crawl a small corpus of parallel dialectal Estonian - standard Estonian sentences. In addition, we take a savvy approach of generating more synthetic training data for the normalization task by using an existing dialect generator model built for Finnish to "dialectalize" standard Estonian sentences from the Universal Dependencies tree banks. Our BERT based normalization model achieves a word error rate that is 26.49 points lower when using both the synthetic data and Estonian data in comparison to training the model with only the available Estonian data. Our results suggest that synthetic data generated by a model trained on a more resourced related language can indeed boost the results for a less resourced language.

AB - While standard Estonian is not a low-resourced language, the different dialects of the language are under-resourced from the point of view of NLP, given that there are no vast hand normalized resources available for training a machine learning model to normalize dialectal Estonian to standard Estonian. In this paper, we crawl a small corpus of parallel dialectal Estonian - standard Estonian sentences. In addition, we take a savvy approach of generating more synthetic training data for the normalization task by using an existing dialect generator model built for Finnish to "dialectalize" standard Estonian sentences from the Universal Dependencies tree banks. Our BERT based normalization model achieves a word error rate that is 26.49 points lower when using both the synthetic data and Estonian data in comparison to training the model with only the available Estonian data. Our results suggest that synthetic data generated by a model trained on a more resourced related language can indeed boost the results for a less resourced language.

UR - https://aclanthology.org/2022.deeplo-1.7.pdf

UR - https://www.scopus.com/pages/publications/85137547883

U2 - 10.18653/v1/2022.deeplo-1.7

DO - 10.18653/v1/2022.deeplo-1.7

M3 - Conference paper

T3 - DeepLo 2022 - 3rd Workshop on Deep Learning Approaches for Low-Resource NLP, Proceedings of the DeepLo Workshop

SP - 61

EP - 66

BT - DeepLo 2022 - 3rd Workshop on Deep Learning Approaches for Low-Resource NLP, Proceedings of the DeepLo Workshop

A2 - Cherry, Colin

A2 - Fan, Angela

A2 - Foster, George

A2 - Haffari, Gholamreza

A2 - Khadivi, Shahram

A2 - Peng, Nanyun

A2 - Ren, Xiang

A2 - Shareghi, Ehsan

A2 - Swayamdipta, Swabha

PB - Association for Computational Linguistics

CY - Seattle

ER -

Tuisk T, Mika H, Khalid A. Help from the Neighbors: Estonian Dialect Normalization Using a Finnish Dialect Generator. (gadā) Cherry C, Fan A, Foster G, Haffari G, Khadivi S, Peng N, Ren X, Shareghi E, Swayamdipta S, redaktori, DeepLo 2022 - 3rd Workshop on Deep Learning Approaches for Low-Resource NLP, Proceedings of the DeepLo Workshop. Seattle: Association for Computational Linguistics. 2022. lpp. 61-66. (DeepLo 2022 - 3rd Workshop on Deep Learning Approaches for Low-Resource NLP, Proceedings of the DeepLo Workshop). doi: 10.18653/v1/2022.deeplo-1.7

Help from the Neighbors: Estonian Dialect Normalization Using a Finnish Dialect Generator

Kopsavilkums

Publikāciju sērijas

OECD Zinātnes nozare

Piekļuve Dokumentiem

Citi faili un saites

Nospiedums

Citēt šo