Mimicking Infants' Bilingual Language Acquisition for Domain Specialized Neural Machine Translation

Chanjun Park; Woo Young Go; Sugyeong Eo; Hyeonseok Moon; Seolhwa Lee; Heuiseok Lim

doi:10.1109/ACCESS.2022.3165572

Mimicking Infants' Bilingual Language Acquisition for Domain Specialized Neural Machine Translation

Chanjun Park, Woo Young Go, Sugyeong Eo, Hyeonseok Moon, Seolhwa Lee, Heuiseok Lim^*

^*Corresponding author af dette arbejde

Natural Language Processing

Publikation: Bidrag til tidsskrift › Tidsskriftartikel › Forskning › peer review

4 Citationer (Scopus)

28 Downloads (Pure)

Abstract

Existing methods of training domain-specialized neural machine translation (DS-NMT) models are based on the pretrain-finetuning approach (PFA). In this study, we reinterpret existing methods based on the perspective of cognitive science related to cross language speech perception. We propose the cross communication method (CCM), a new DS-NMT training approach. Inspired by the learning method of infants, we perform DS-NMT training by configuring and training DC and GC concurrently in batches. Quantitative and qualitative analysis of our experimental results show that CCM can achieve superior performance compared to the conventional methods. Additionally, we conducted an experiment considering the DS-NMT service to meet industrial demands.

Originalsprog	Engelsk
Tidsskrift	IEEE Access
Vol/bind	10
Sider (fra-til)	38684-38693
ISSN	2169-3536
DOI	https://doi.org/10.1109/ACCESS.2022.3165572
Status	Udgivet - 2022

Bibliografisk note

Funding Information:
This work was supported in part by the Ministry of Science and ICT (MSIT), South Korea, through the Information Technology Research Center (ITRC) Support Program supervised by the Institute of Information and Communications Technology Planning and Evaluation (IITP) under Grant IITP-2018-0-01405; in part by the IITP Grant funded by the Korea Government (MSIT) (A Neural-Symbolic Model for Knowledge Acquisition and Inference Techniques) under Grant 2020-0-00368; and in part by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education under Grant NRF-2021R1A6A1A03045425

Publisher Copyright:
© 2013 IEEE.

Adgang til dokumentet

10.1109/ACCESS.2022.3165572Licens: CC BY

FulltextForlagets udgivne version, 1,37 MBLicens: CC BY

Biblioteksadgang?

Tjek tilgængelighed hos det Kgl. Bibliotek

Andre filer og links

Link to publication in Scopus

Citationsformater

@article{94ffc31588704f988756a29f50530981,

title = "Mimicking Infants' Bilingual Language Acquisition for Domain Specialized Neural Machine Translation",

abstract = "Existing methods of training domain-specialized neural machine translation (DS-NMT) models are based on the pretrain-finetuning approach (PFA). In this study, we reinterpret existing methods based on the perspective of cognitive science related to cross language speech perception. We propose the cross communication method (CCM), a new DS-NMT training approach. Inspired by the learning method of infants, we perform DS-NMT training by configuring and training DC and GC concurrently in batches. Quantitative and qualitative analysis of our experimental results show that CCM can achieve superior performance compared to the conventional methods. Additionally, we conducted an experiment considering the DS-NMT service to meet industrial demands. ",

keywords = "cross communication method, deep learning, Domain-specialized neural machine translation, neural machine translation",

author = "Chanjun Park and Go, {Woo Young} and Sugyeong Eo and Hyeonseok Moon and Seolhwa Lee and Heuiseok Lim",

note = "Publisher Copyright: {\textcopyright} 2013 IEEE.",

year = "2022",

doi = "10.1109/ACCESS.2022.3165572",

language = "English",

volume = "10",

pages = "38684--38693",

journal = "IEEE Access",

issn = "2169-3536",

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

}

TY - JOUR

T1 - Mimicking Infants' Bilingual Language Acquisition for Domain Specialized Neural Machine Translation

AU - Park, Chanjun

AU - Go, Woo Young

AU - Eo, Sugyeong

AU - Moon, Hyeonseok

AU - Lee, Seolhwa

AU - Lim, Heuiseok

PY - 2022

Y1 - 2022

N2 - Existing methods of training domain-specialized neural machine translation (DS-NMT) models are based on the pretrain-finetuning approach (PFA). In this study, we reinterpret existing methods based on the perspective of cognitive science related to cross language speech perception. We propose the cross communication method (CCM), a new DS-NMT training approach. Inspired by the learning method of infants, we perform DS-NMT training by configuring and training DC and GC concurrently in batches. Quantitative and qualitative analysis of our experimental results show that CCM can achieve superior performance compared to the conventional methods. Additionally, we conducted an experiment considering the DS-NMT service to meet industrial demands.

AB - Existing methods of training domain-specialized neural machine translation (DS-NMT) models are based on the pretrain-finetuning approach (PFA). In this study, we reinterpret existing methods based on the perspective of cognitive science related to cross language speech perception. We propose the cross communication method (CCM), a new DS-NMT training approach. Inspired by the learning method of infants, we perform DS-NMT training by configuring and training DC and GC concurrently in batches. Quantitative and qualitative analysis of our experimental results show that CCM can achieve superior performance compared to the conventional methods. Additionally, we conducted an experiment considering the DS-NMT service to meet industrial demands.

KW - cross communication method

KW - deep learning

KW - Domain-specialized neural machine translation

KW - neural machine translation

UR - http://www.scopus.com/inward/record.url?scp=85128258171&partnerID=8YFLogxK

U2 - 10.1109/ACCESS.2022.3165572

DO - 10.1109/ACCESS.2022.3165572

M3 - Journal article

AN - SCOPUS:85128258171

SN - 2169-3536

VL - 10

SP - 38684

EP - 38693

JO - IEEE Access

JF - IEEE Access

ER -