Unsupervised Domain Adaptation Based on Style Aware

In recent years,neural machine translation has made significant progress in translation quality,but it relies on parallel bilingual sentence pairs heavily during the training process.However,parallel resources are scarce for the e-commerce domain,in addition,cultural differences lead to stylistic di...

Full description

Bibliographic Details
Main Author: NING Qiu-yi, SHI Xiao-jing, DUAN Xiang-yu, ZHANG Min
Format: Article
Language:zho
Published: Editorial office of Computer Science 2022-01-01
Series:Jisuanji kexue
Subjects:
Online Access:https://www.jsjkx.com/fileup/1002-137X/PDF/1002-137X-2022-1-271.pdf
_version_ 1818157912668766208
author NING Qiu-yi, SHI Xiao-jing, DUAN Xiang-yu, ZHANG Min
author_facet NING Qiu-yi, SHI Xiao-jing, DUAN Xiang-yu, ZHANG Min
author_sort NING Qiu-yi, SHI Xiao-jing, DUAN Xiang-yu, ZHANG Min
collection DOAJ
description In recent years,neural machine translation has made significant progress in translation quality,but it relies on parallel bilingual sentence pairs heavily during the training process.However,parallel resources are scarce for the e-commerce domain,in addition,cultural differences lead to stylistic differences in product information expression.To solve these two problems,a style-aware unsupervised domain adaptation algorithm is proposed,which makes full use of e-commerce monolingual data in the mutual training method,while introducing quasi knowledge distillation approach to deal with style differences.We construct non-parallel bilingual corpus by obtaining e-commerce product data information,and then carry out experiments based on the aforementioned corpus and Chinese and English news parallel corpus.The results show that the algorithm significantly improves translation qua-lity compared to various unsupervised domain adaptation methods,improves about 5 BLEU points compared with the strongest baseline system.In addition,the algorithm is further extended to Ted,Law and Medical OPUS data,all of which achieve better translation results.
first_indexed 2024-12-11T15:21:44Z
format Article
id doaj.art-6e0edb1eaee84235bedc22276201dc65
institution Directory Open Access Journal
issn 1002-137X
language zho
last_indexed 2024-12-11T15:21:44Z
publishDate 2022-01-01
publisher Editorial office of Computer Science
record_format Article
series Jisuanji kexue
spelling doaj.art-6e0edb1eaee84235bedc22276201dc652022-12-22T01:00:21ZzhoEditorial office of Computer ScienceJisuanji kexue1002-137X2022-01-0149127127810.11896/jsjkx.201200094Unsupervised Domain Adaptation Based on Style AwareNING Qiu-yi, SHI Xiao-jing, DUAN Xiang-yu, ZHANG Min0School of Computer Science and Technology,Soochow University,Suzhou,Jiangsu 215006,ChinaIn recent years,neural machine translation has made significant progress in translation quality,but it relies on parallel bilingual sentence pairs heavily during the training process.However,parallel resources are scarce for the e-commerce domain,in addition,cultural differences lead to stylistic differences in product information expression.To solve these two problems,a style-aware unsupervised domain adaptation algorithm is proposed,which makes full use of e-commerce monolingual data in the mutual training method,while introducing quasi knowledge distillation approach to deal with style differences.We construct non-parallel bilingual corpus by obtaining e-commerce product data information,and then carry out experiments based on the aforementioned corpus and Chinese and English news parallel corpus.The results show that the algorithm significantly improves translation qua-lity compared to various unsupervised domain adaptation methods,improves about 5 BLEU points compared with the strongest baseline system.In addition,the algorithm is further extended to Ted,Law and Medical OPUS data,all of which achieve better translation results.https://www.jsjkx.com/fileup/1002-137X/PDF/1002-137X-2022-1-271.pdfmachine translation|unsupervised|domain adaptation|style aware|e-commerce
spellingShingle NING Qiu-yi, SHI Xiao-jing, DUAN Xiang-yu, ZHANG Min
Unsupervised Domain Adaptation Based on Style Aware
Jisuanji kexue
machine translation|unsupervised|domain adaptation|style aware|e-commerce
title Unsupervised Domain Adaptation Based on Style Aware
title_full Unsupervised Domain Adaptation Based on Style Aware
title_fullStr Unsupervised Domain Adaptation Based on Style Aware
title_full_unstemmed Unsupervised Domain Adaptation Based on Style Aware
title_short Unsupervised Domain Adaptation Based on Style Aware
title_sort unsupervised domain adaptation based on style aware
topic machine translation|unsupervised|domain adaptation|style aware|e-commerce
url https://www.jsjkx.com/fileup/1002-137X/PDF/1002-137X-2022-1-271.pdf
work_keys_str_mv AT ningqiuyishixiaojingduanxiangyuzhangmin unsuperviseddomainadaptationbasedonstyleaware