The Effect of Normalization for Bi-directional Amharic-English Neural Machine Translation

Tadesse Destaw Belay; Atnafu Lambebo Tonja; Olga Kolesnikova; Seid Muhie Yimam; Abinew Ali Ayele; Silesh Bogale Haile; Grigori Sidorov; Alexander Gelbukh

doi:10.1109/ICT4DA56482.2022.9971385

The Effect of Normalization for Bi-directional Amharic-English Neural Machine Translation

Tadesse Destaw Belay, Atnafu Lambebo Tonja, Olga Kolesnikova, Seid Muhie Yimam, Abinew Ali Ayele, Silesh Bogale Haile, Grigori Sidorov, Alexander Gelbukh

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

3 Scopus citations

Abstract

Machine translation (MT) is one of the prominent tasks in natural language processing whose objective is to translate texts automatically from one natural language to another. Nowadays, using deep neural networks for MT task has received a great deal of attention. These networks require lots of data to learn abstract representations of the input and store it in continuous vectors. This paper presents the first relatively large-scale Amharic-English parallel sentence dataset. Using these compiled data, we build bi-directional Amharic-English translation models by fine-tuning the existing Facebook M2M100 pre-trained model achieving a BLEU score of 37.79 in Amharic-English translation and 32.74 in English-Amharic translation. Additionally, we explore the effects of Amharic homophone normalization on the machine translation task. The results show that normalization of Amharic homophone characters increases the performance of Amharic-English machine translation in both directions.

Original language	English
Title of host publication	2022 International Conference on Information and Communication Technology for Development for Africa, ICT4DA 2022
Editors	Esubalew Alemneh, Ethiopia Nigussie, Fisseha Mekuria
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	84-89
Number of pages	6
ISBN (Electronic)	9781665455879
DOIs	https://doi.org/10.1109/ICT4DA56482.2022.9971385
State	Published - 2022
Event	2022 International Conference on Information and Communication Technology for Development for Africa, ICT4DA 2022 - Bahir Dar, Ethiopia Duration: 28 Nov 2022 → 30 Nov 2022

Publication series

Name	2022 International Conference on Information and Communication Technology for Development for Africa, ICT4DA 2022

Conference

Conference	2022 International Conference on Information and Communication Technology for Development for Africa, ICT4DA 2022
Country/Territory	Ethiopia
City	Bahir Dar
Period	28/11/22 → 30/11/22

Keywords

Amharic-English MT
Neural machine translation
homophone normalization
low-resourced language
pre-trained models

Access to Document

10.1109/ICT4DA56482.2022.9971385

Cite this

Belay, T. D., Tonja, A. L., Kolesnikova, O., Yimam, S. M., Ayele, A. A., Haile, S. B., Sidorov, G., & Gelbukh, A. (2022). The Effect of Normalization for Bi-directional Amharic-English Neural Machine Translation. In E. Alemneh, E. Nigussie, & F. Mekuria (Eds.), 2022 International Conference on Information and Communication Technology for Development for Africa, ICT4DA 2022 (pp. 84-89). (2022 International Conference on Information and Communication Technology for Development for Africa, ICT4DA 2022). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ICT4DA56482.2022.9971385

Belay, Tadesse Destaw ; Tonja, Atnafu Lambebo ; Kolesnikova, Olga et al. / The Effect of Normalization for Bi-directional Amharic-English Neural Machine Translation. 2022 International Conference on Information and Communication Technology for Development for Africa, ICT4DA 2022. editor / Esubalew Alemneh ; Ethiopia Nigussie ; Fisseha Mekuria. Institute of Electrical and Electronics Engineers Inc., 2022. pp. 84-89 (2022 International Conference on Information and Communication Technology for Development for Africa, ICT4DA 2022).

@inproceedings{e656c3c34127415b95195787b9947064,

title = "The Effect of Normalization for Bi-directional Amharic-English Neural Machine Translation",

abstract = "Machine translation (MT) is one of the prominent tasks in natural language processing whose objective is to translate texts automatically from one natural language to another. Nowadays, using deep neural networks for MT task has received a great deal of attention. These networks require lots of data to learn abstract representations of the input and store it in continuous vectors. This paper presents the first relatively large-scale Amharic-English parallel sentence dataset. Using these compiled data, we build bi-directional Amharic-English translation models by fine-tuning the existing Facebook M2M100 pre-trained model achieving a BLEU score of 37.79 in Amharic-English translation and 32.74 in English-Amharic translation. Additionally, we explore the effects of Amharic homophone normalization on the machine translation task. The results show that normalization of Amharic homophone characters increases the performance of Amharic-English machine translation in both directions.",

keywords = "Amharic-English MT, Neural machine translation, homophone normalization, low-resourced language, pre-trained models",

author = "Belay, {Tadesse Destaw} and Tonja, {Atnafu Lambebo} and Olga Kolesnikova and Yimam, {Seid Muhie} and Ayele, {Abinew Ali} and Haile, {Silesh Bogale} and Grigori Sidorov and Alexander Gelbukh",

note = "Publisher Copyright: {\textcopyright} 2022 IEEE.; 2022 International Conference on Information and Communication Technology for Development for Africa, ICT4DA 2022 ; Conference date: 28-11-2022 Through 30-11-2022",

year = "2022",

doi = "10.1109/ICT4DA56482.2022.9971385",

language = "Ingl{\'e}s",

series = "2022 International Conference on Information and Communication Technology for Development for Africa, ICT4DA 2022",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "84--89",

editor = "Esubalew Alemneh and Ethiopia Nigussie and Fisseha Mekuria",

booktitle = "2022 International Conference on Information and Communication Technology for Development for Africa, ICT4DA 2022",

address = "Estados Unidos",

}

Belay, TD, Tonja, AL, Kolesnikova, O, Yimam, SM, Ayele, AA, Haile, SB, Sidorov, G & Gelbukh, A 2022, The Effect of Normalization for Bi-directional Amharic-English Neural Machine Translation. in E Alemneh, E Nigussie & F Mekuria (eds), 2022 International Conference on Information and Communication Technology for Development for Africa, ICT4DA 2022. 2022 International Conference on Information and Communication Technology for Development for Africa, ICT4DA 2022, Institute of Electrical and Electronics Engineers Inc., pp. 84-89, 2022 International Conference on Information and Communication Technology for Development for Africa, ICT4DA 2022, Bahir Dar, Ethiopia, 28/11/22. https://doi.org/10.1109/ICT4DA56482.2022.9971385

The Effect of Normalization for Bi-directional Amharic-English Neural Machine Translation. / Belay, Tadesse Destaw; Tonja, Atnafu Lambebo; Kolesnikova, Olga et al.
2022 International Conference on Information and Communication Technology for Development for Africa, ICT4DA 2022. ed. / Esubalew Alemneh; Ethiopia Nigussie; Fisseha Mekuria. Institute of Electrical and Electronics Engineers Inc., 2022. p. 84-89 (2022 International Conference on Information and Communication Technology for Development for Africa, ICT4DA 2022).

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

TY - GEN

T1 - The Effect of Normalization for Bi-directional Amharic-English Neural Machine Translation

AU - Belay, Tadesse Destaw

AU - Tonja, Atnafu Lambebo

AU - Kolesnikova, Olga

AU - Yimam, Seid Muhie

AU - Ayele, Abinew Ali

AU - Haile, Silesh Bogale

AU - Sidorov, Grigori

AU - Gelbukh, Alexander

PY - 2022

Y1 - 2022

N2 - Machine translation (MT) is one of the prominent tasks in natural language processing whose objective is to translate texts automatically from one natural language to another. Nowadays, using deep neural networks for MT task has received a great deal of attention. These networks require lots of data to learn abstract representations of the input and store it in continuous vectors. This paper presents the first relatively large-scale Amharic-English parallel sentence dataset. Using these compiled data, we build bi-directional Amharic-English translation models by fine-tuning the existing Facebook M2M100 pre-trained model achieving a BLEU score of 37.79 in Amharic-English translation and 32.74 in English-Amharic translation. Additionally, we explore the effects of Amharic homophone normalization on the machine translation task. The results show that normalization of Amharic homophone characters increases the performance of Amharic-English machine translation in both directions.

AB - Machine translation (MT) is one of the prominent tasks in natural language processing whose objective is to translate texts automatically from one natural language to another. Nowadays, using deep neural networks for MT task has received a great deal of attention. These networks require lots of data to learn abstract representations of the input and store it in continuous vectors. This paper presents the first relatively large-scale Amharic-English parallel sentence dataset. Using these compiled data, we build bi-directional Amharic-English translation models by fine-tuning the existing Facebook M2M100 pre-trained model achieving a BLEU score of 37.79 in Amharic-English translation and 32.74 in English-Amharic translation. Additionally, we explore the effects of Amharic homophone normalization on the machine translation task. The results show that normalization of Amharic homophone characters increases the performance of Amharic-English machine translation in both directions.

KW - Amharic-English MT

KW - Neural machine translation

KW - homophone normalization

KW - low-resourced language

KW - pre-trained models

UR - http://www.scopus.com/inward/record.url?scp=85145438495&partnerID=8YFLogxK

U2 - 10.1109/ICT4DA56482.2022.9971385

DO - 10.1109/ICT4DA56482.2022.9971385

M3 - Contribución a la conferencia

AN - SCOPUS:85145438495

T3 - 2022 International Conference on Information and Communication Technology for Development for Africa, ICT4DA 2022

SP - 84

EP - 89

BT - 2022 International Conference on Information and Communication Technology for Development for Africa, ICT4DA 2022

A2 - Alemneh, Esubalew

A2 - Nigussie, Ethiopia

A2 - Mekuria, Fisseha

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2022 International Conference on Information and Communication Technology for Development for Africa, ICT4DA 2022

Y2 - 28 November 2022 through 30 November 2022

ER -

Belay TD, Tonja AL, Kolesnikova O, Yimam SM, Ayele AA, Haile SB et al. The Effect of Normalization for Bi-directional Amharic-English Neural Machine Translation. In Alemneh E, Nigussie E, Mekuria F, editors, 2022 International Conference on Information and Communication Technology for Development for Africa, ICT4DA 2022. Institute of Electrical and Electronics Engineers Inc. 2022. p. 84-89. (2022 International Conference on Information and Communication Technology for Development for Africa, ICT4DA 2022). doi: 10.1109/ICT4DA56482.2022.9971385

The Effect of Normalization for Bi-directional Amharic-English Neural Machine Translation

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this