Loading…

LLMs-based machine translation for E-commerce

Large language models(LLMs) have shown promising performance for various downstream tasks, especially machine translation. However, LLMs and Specialized Translation Models (STMs) are designed to handle general translation needs, they are not well-suited for domains with specialized terms and writing...

Full description

Saved in:
Bibliographic Details
Published in:Expert systems with applications 2024-12, Vol.258, p.125087, Article 125087
Main Authors: Gao, Dehong, Chen, Kaidi, Chen, Ben, Dai, Huangyu, Jin, Linbo, Jiang, Wen, Ning, Wei, Yu, Shanqing, Xuan, Qi, Cai, Xiaoyan, Yang, Libin, Wang, Zhen
Format: Article
Language:English
Subjects:
Citations: Items that this one cites
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
cited_by
cites cdi_FETCH-LOGICAL-c181t-4e1e48ad4bb9e4ae913997e9e742265dfb7706f7fe3fa2d17061fa8a420bce23
container_end_page
container_issue
container_start_page 125087
container_title Expert systems with applications
container_volume 258
creator Gao, Dehong
Chen, Kaidi
Chen, Ben
Dai, Huangyu
Jin, Linbo
Jiang, Wen
Ning, Wei
Yu, Shanqing
Xuan, Qi
Cai, Xiaoyan
Yang, Libin
Wang, Zhen
description Large language models(LLMs) have shown promising performance for various downstream tasks, especially machine translation. However, LLMs and Specialized Translation Models (STMs) are designed to handle general translation needs, they are not well-suited for domains with specialized terms and writing styles, such as e-commerce, legal, and medicine. In the e-commerce domain, the text often contains many domain-specific terms and keyword-stacked structures, leading to poor translation quality with existing NMT methods. To tackle these problems, we have collected two resources specifically for the e-commerce domain, including aligned Chinese-English bilingual terms and parallel corpus from real e-commerce scenarios for model fine-tuning. We propose an LLMs-based E-commerce machine translation approach(LEMT) which includes LLMs utilization, e-commerce resources collection, and tokenizer optimization. We conduct two-stage fine-tuning and self-contrastive enhancement based on general LLMs to enable the model to learn translation features in the e-commerce domain. Through comprehensive evaluations on real e-commerce titles, our LEMT methodology demonstrates superior translation quality and robustness, outperforming leading NMT models such as NLLB, LLaMA, and even GPT-4. •Existing universal translation models perform poorly in the e-commerce field.•Translation models based on LLMs improve translation capabilities more efficiently.•Resources based on e-commerce texts are conducive to e-commerce translation.•Two-step fine-tuning gradually enhance the translation ability of specialized domain.
doi_str_mv 10.1016/j.eswa.2024.125087
format article
fullrecord <record><control><sourceid>elsevier_cross</sourceid><recordid>TN_cdi_crossref_primary_10_1016_j_eswa_2024_125087</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><els_id>S0957417424019547</els_id><sourcerecordid>S0957417424019547</sourcerecordid><originalsourceid>FETCH-LOGICAL-c181t-4e1e48ad4bb9e4ae913997e9e742265dfb7706f7fe3fa2d17061fa8a420bce23</originalsourceid><addsrcrecordid>eNp9j81KxDAURrNQcBx9AVd9gdR707RpwY0M4w9U3Mw-pMkNpkxbSYri29tS166-1fk4h7E7hBwBq_s-p_RtcgFC5ihKqNUF20FTKi5RySt2nVIPgApA7Rhv27fEO5PIZYOxH2GkbI5mTGczh2nM_BSzI7fTMFC0dMMuvTknuv3bPTs9HU-HF96-P78eHltuscaZS0KStXGy6xqShhosmkZRQ0oKUZXOd0pB5ZWnwhvhFpUKvamNFNBZEsWeie3WximlSF5_xjCY-KMR9Nqoe7026rVRb40L9LBBtIh9BYo62UCjJRci2Vm7KfyH_wLtOFwS</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>LLMs-based machine translation for E-commerce</title><source>ScienceDirect Freedom Collection 2022-2024</source><creator>Gao, Dehong ; Chen, Kaidi ; Chen, Ben ; Dai, Huangyu ; Jin, Linbo ; Jiang, Wen ; Ning, Wei ; Yu, Shanqing ; Xuan, Qi ; Cai, Xiaoyan ; Yang, Libin ; Wang, Zhen</creator><creatorcontrib>Gao, Dehong ; Chen, Kaidi ; Chen, Ben ; Dai, Huangyu ; Jin, Linbo ; Jiang, Wen ; Ning, Wei ; Yu, Shanqing ; Xuan, Qi ; Cai, Xiaoyan ; Yang, Libin ; Wang, Zhen</creatorcontrib><description>Large language models(LLMs) have shown promising performance for various downstream tasks, especially machine translation. However, LLMs and Specialized Translation Models (STMs) are designed to handle general translation needs, they are not well-suited for domains with specialized terms and writing styles, such as e-commerce, legal, and medicine. In the e-commerce domain, the text often contains many domain-specific terms and keyword-stacked structures, leading to poor translation quality with existing NMT methods. To tackle these problems, we have collected two resources specifically for the e-commerce domain, including aligned Chinese-English bilingual terms and parallel corpus from real e-commerce scenarios for model fine-tuning. We propose an LLMs-based E-commerce machine translation approach(LEMT) which includes LLMs utilization, e-commerce resources collection, and tokenizer optimization. We conduct two-stage fine-tuning and self-contrastive enhancement based on general LLMs to enable the model to learn translation features in the e-commerce domain. Through comprehensive evaluations on real e-commerce titles, our LEMT methodology demonstrates superior translation quality and robustness, outperforming leading NMT models such as NLLB, LLaMA, and even GPT-4. •Existing universal translation models perform poorly in the e-commerce field.•Translation models based on LLMs improve translation capabilities more efficiently.•Resources based on e-commerce texts are conducive to e-commerce translation.•Two-step fine-tuning gradually enhance the translation ability of specialized domain.</description><identifier>ISSN: 0957-4174</identifier><identifier>DOI: 10.1016/j.eswa.2024.125087</identifier><language>eng</language><publisher>Elsevier Ltd</publisher><subject>E-commerce domain ; Fine-tuning ; Large language models ; Neural machine translation ; Self-contrastive</subject><ispartof>Expert systems with applications, 2024-12, Vol.258, p.125087, Article 125087</ispartof><rights>2024</rights><lds50>peer_reviewed</lds50><woscitedreferencessubscribed>false</woscitedreferencessubscribed><cites>FETCH-LOGICAL-c181t-4e1e48ad4bb9e4ae913997e9e742265dfb7706f7fe3fa2d17061fa8a420bce23</cites><orcidid>0000-0003-4495-8686 ; 0000-0002-3844-8359 ; 0000-0002-6320-7012 ; 0000-0001-8249-6879 ; 0000-0003-3201-2113 ; 0009-0008-9688-5101 ; 0000-0001-8487-9910 ; 0000-0002-6636-5702 ; 0000-0001-5170-8082 ; 0000-0001-5316-7689 ; 0000-0002-8182-2852 ; 0000-0002-1406-107X</orcidid></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>314,778,782,27907,27908</link.rule.ids></links><search><creatorcontrib>Gao, Dehong</creatorcontrib><creatorcontrib>Chen, Kaidi</creatorcontrib><creatorcontrib>Chen, Ben</creatorcontrib><creatorcontrib>Dai, Huangyu</creatorcontrib><creatorcontrib>Jin, Linbo</creatorcontrib><creatorcontrib>Jiang, Wen</creatorcontrib><creatorcontrib>Ning, Wei</creatorcontrib><creatorcontrib>Yu, Shanqing</creatorcontrib><creatorcontrib>Xuan, Qi</creatorcontrib><creatorcontrib>Cai, Xiaoyan</creatorcontrib><creatorcontrib>Yang, Libin</creatorcontrib><creatorcontrib>Wang, Zhen</creatorcontrib><title>LLMs-based machine translation for E-commerce</title><title>Expert systems with applications</title><description>Large language models(LLMs) have shown promising performance for various downstream tasks, especially machine translation. However, LLMs and Specialized Translation Models (STMs) are designed to handle general translation needs, they are not well-suited for domains with specialized terms and writing styles, such as e-commerce, legal, and medicine. In the e-commerce domain, the text often contains many domain-specific terms and keyword-stacked structures, leading to poor translation quality with existing NMT methods. To tackle these problems, we have collected two resources specifically for the e-commerce domain, including aligned Chinese-English bilingual terms and parallel corpus from real e-commerce scenarios for model fine-tuning. We propose an LLMs-based E-commerce machine translation approach(LEMT) which includes LLMs utilization, e-commerce resources collection, and tokenizer optimization. We conduct two-stage fine-tuning and self-contrastive enhancement based on general LLMs to enable the model to learn translation features in the e-commerce domain. Through comprehensive evaluations on real e-commerce titles, our LEMT methodology demonstrates superior translation quality and robustness, outperforming leading NMT models such as NLLB, LLaMA, and even GPT-4. •Existing universal translation models perform poorly in the e-commerce field.•Translation models based on LLMs improve translation capabilities more efficiently.•Resources based on e-commerce texts are conducive to e-commerce translation.•Two-step fine-tuning gradually enhance the translation ability of specialized domain.</description><subject>E-commerce domain</subject><subject>Fine-tuning</subject><subject>Large language models</subject><subject>Neural machine translation</subject><subject>Self-contrastive</subject><issn>0957-4174</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2024</creationdate><recordtype>article</recordtype><recordid>eNp9j81KxDAURrNQcBx9AVd9gdR707RpwY0M4w9U3Mw-pMkNpkxbSYri29tS166-1fk4h7E7hBwBq_s-p_RtcgFC5ihKqNUF20FTKi5RySt2nVIPgApA7Rhv27fEO5PIZYOxH2GkbI5mTGczh2nM_BSzI7fTMFC0dMMuvTknuv3bPTs9HU-HF96-P78eHltuscaZS0KStXGy6xqShhosmkZRQ0oKUZXOd0pB5ZWnwhvhFpUKvamNFNBZEsWeie3WximlSF5_xjCY-KMR9Nqoe7026rVRb40L9LBBtIh9BYo62UCjJRci2Vm7KfyH_wLtOFwS</recordid><startdate>20241215</startdate><enddate>20241215</enddate><creator>Gao, Dehong</creator><creator>Chen, Kaidi</creator><creator>Chen, Ben</creator><creator>Dai, Huangyu</creator><creator>Jin, Linbo</creator><creator>Jiang, Wen</creator><creator>Ning, Wei</creator><creator>Yu, Shanqing</creator><creator>Xuan, Qi</creator><creator>Cai, Xiaoyan</creator><creator>Yang, Libin</creator><creator>Wang, Zhen</creator><general>Elsevier Ltd</general><scope>AAYXX</scope><scope>CITATION</scope><orcidid>https://orcid.org/0000-0003-4495-8686</orcidid><orcidid>https://orcid.org/0000-0002-3844-8359</orcidid><orcidid>https://orcid.org/0000-0002-6320-7012</orcidid><orcidid>https://orcid.org/0000-0001-8249-6879</orcidid><orcidid>https://orcid.org/0000-0003-3201-2113</orcidid><orcidid>https://orcid.org/0009-0008-9688-5101</orcidid><orcidid>https://orcid.org/0000-0001-8487-9910</orcidid><orcidid>https://orcid.org/0000-0002-6636-5702</orcidid><orcidid>https://orcid.org/0000-0001-5170-8082</orcidid><orcidid>https://orcid.org/0000-0001-5316-7689</orcidid><orcidid>https://orcid.org/0000-0002-8182-2852</orcidid><orcidid>https://orcid.org/0000-0002-1406-107X</orcidid></search><sort><creationdate>20241215</creationdate><title>LLMs-based machine translation for E-commerce</title><author>Gao, Dehong ; Chen, Kaidi ; Chen, Ben ; Dai, Huangyu ; Jin, Linbo ; Jiang, Wen ; Ning, Wei ; Yu, Shanqing ; Xuan, Qi ; Cai, Xiaoyan ; Yang, Libin ; Wang, Zhen</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c181t-4e1e48ad4bb9e4ae913997e9e742265dfb7706f7fe3fa2d17061fa8a420bce23</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2024</creationdate><topic>E-commerce domain</topic><topic>Fine-tuning</topic><topic>Large language models</topic><topic>Neural machine translation</topic><topic>Self-contrastive</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Gao, Dehong</creatorcontrib><creatorcontrib>Chen, Kaidi</creatorcontrib><creatorcontrib>Chen, Ben</creatorcontrib><creatorcontrib>Dai, Huangyu</creatorcontrib><creatorcontrib>Jin, Linbo</creatorcontrib><creatorcontrib>Jiang, Wen</creatorcontrib><creatorcontrib>Ning, Wei</creatorcontrib><creatorcontrib>Yu, Shanqing</creatorcontrib><creatorcontrib>Xuan, Qi</creatorcontrib><creatorcontrib>Cai, Xiaoyan</creatorcontrib><creatorcontrib>Yang, Libin</creatorcontrib><creatorcontrib>Wang, Zhen</creatorcontrib><collection>CrossRef</collection><jtitle>Expert systems with applications</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Gao, Dehong</au><au>Chen, Kaidi</au><au>Chen, Ben</au><au>Dai, Huangyu</au><au>Jin, Linbo</au><au>Jiang, Wen</au><au>Ning, Wei</au><au>Yu, Shanqing</au><au>Xuan, Qi</au><au>Cai, Xiaoyan</au><au>Yang, Libin</au><au>Wang, Zhen</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>LLMs-based machine translation for E-commerce</atitle><jtitle>Expert systems with applications</jtitle><date>2024-12-15</date><risdate>2024</risdate><volume>258</volume><spage>125087</spage><pages>125087-</pages><artnum>125087</artnum><issn>0957-4174</issn><abstract>Large language models(LLMs) have shown promising performance for various downstream tasks, especially machine translation. However, LLMs and Specialized Translation Models (STMs) are designed to handle general translation needs, they are not well-suited for domains with specialized terms and writing styles, such as e-commerce, legal, and medicine. In the e-commerce domain, the text often contains many domain-specific terms and keyword-stacked structures, leading to poor translation quality with existing NMT methods. To tackle these problems, we have collected two resources specifically for the e-commerce domain, including aligned Chinese-English bilingual terms and parallel corpus from real e-commerce scenarios for model fine-tuning. We propose an LLMs-based E-commerce machine translation approach(LEMT) which includes LLMs utilization, e-commerce resources collection, and tokenizer optimization. We conduct two-stage fine-tuning and self-contrastive enhancement based on general LLMs to enable the model to learn translation features in the e-commerce domain. Through comprehensive evaluations on real e-commerce titles, our LEMT methodology demonstrates superior translation quality and robustness, outperforming leading NMT models such as NLLB, LLaMA, and even GPT-4. •Existing universal translation models perform poorly in the e-commerce field.•Translation models based on LLMs improve translation capabilities more efficiently.•Resources based on e-commerce texts are conducive to e-commerce translation.•Two-step fine-tuning gradually enhance the translation ability of specialized domain.</abstract><pub>Elsevier Ltd</pub><doi>10.1016/j.eswa.2024.125087</doi><orcidid>https://orcid.org/0000-0003-4495-8686</orcidid><orcidid>https://orcid.org/0000-0002-3844-8359</orcidid><orcidid>https://orcid.org/0000-0002-6320-7012</orcidid><orcidid>https://orcid.org/0000-0001-8249-6879</orcidid><orcidid>https://orcid.org/0000-0003-3201-2113</orcidid><orcidid>https://orcid.org/0009-0008-9688-5101</orcidid><orcidid>https://orcid.org/0000-0001-8487-9910</orcidid><orcidid>https://orcid.org/0000-0002-6636-5702</orcidid><orcidid>https://orcid.org/0000-0001-5170-8082</orcidid><orcidid>https://orcid.org/0000-0001-5316-7689</orcidid><orcidid>https://orcid.org/0000-0002-8182-2852</orcidid><orcidid>https://orcid.org/0000-0002-1406-107X</orcidid></addata></record>
fulltext fulltext
identifier ISSN: 0957-4174
ispartof Expert systems with applications, 2024-12, Vol.258, p.125087, Article 125087
issn 0957-4174
language eng
recordid cdi_crossref_primary_10_1016_j_eswa_2024_125087
source ScienceDirect Freedom Collection 2022-2024
subjects E-commerce domain
Fine-tuning
Large language models
Neural machine translation
Self-contrastive
title LLMs-based machine translation for E-commerce
url http://sfxeu10.hosted.exlibrisgroup.com/loughborough?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-16T23%3A30%3A21IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-elsevier_cross&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=LLMs-based%20machine%20translation%20for%20E-commerce&rft.jtitle=Expert%20systems%20with%20applications&rft.au=Gao,%20Dehong&rft.date=2024-12-15&rft.volume=258&rft.spage=125087&rft.pages=125087-&rft.artnum=125087&rft.issn=0957-4174&rft_id=info:doi/10.1016/j.eswa.2024.125087&rft_dat=%3Celsevier_cross%3ES0957417424019547%3C/elsevier_cross%3E%3Cgrp_id%3Ecdi_FETCH-LOGICAL-c181t-4e1e48ad4bb9e4ae913997e9e742265dfb7706f7fe3fa2d17061fa8a420bce23%3C/grp_id%3E%3Coa%3E%3C/oa%3E%3Curl%3E%3C/url%3E&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true