Revisiting the Markov Property for machine translation

In this paper, we re-examine the Markov property in the context of neural machine translation. We design a Markov Autoregressive Transformer (MAT) and undertake a comprehensive assessment of its performance across four WMT benchmarks. Our findings indicate that MAT with an order larger than 4 can ge...

Full description

Saved in:
Bibliographic Details
Main Authors: DU, Cunxiao, ZHOU, Hao, TU, Zhaopeng, JIANG, Jing
Format: text
Language:English
Published: Institutional Knowledge at Singapore Management University 2024
Subjects:
Online Access:https://ink.library.smu.edu.sg/sis_research/8725
https://ink.library.smu.edu.sg/context/sis_research/article/9728/viewcontent/2024.findings_eacl.40_pvoa.pdf
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Singapore Management University
Language: English
id sg-smu-ink.sis_research-9728
record_format dspace
spelling sg-smu-ink.sis_research-97282024-04-18T07:35:08Z Revisiting the Markov Property for machine translation DU, Cunxiao ZHOU, Hao TU, Zhaopeng JIANG, Jing In this paper, we re-examine the Markov property in the context of neural machine translation. We design a Markov Autoregressive Transformer (MAT) and undertake a comprehensive assessment of its performance across four WMT benchmarks. Our findings indicate that MAT with an order larger than 4 can generate translations with quality on par with that of conventional autoregressive transformers. In addition, counter-intuitively, we also find that the advantages of utilizing a higher-order MAT do not specifically contribute to the translation of longer sentences. 2024-03-01T08:00:00Z text application/pdf https://ink.library.smu.edu.sg/sis_research/8725 https://ink.library.smu.edu.sg/context/sis_research/article/9728/viewcontent/2024.findings_eacl.40_pvoa.pdf http://creativecommons.org/licenses/by-nc-nd/4.0/ Research Collection School Of Computing and Information Systems eng Institutional Knowledge at Singapore Management University Numerical Analysis and Scientific Computing
institution Singapore Management University
building SMU Libraries
continent Asia
country Singapore
Singapore
content_provider SMU Libraries
collection InK@SMU
language English
topic Numerical Analysis and Scientific Computing
spellingShingle Numerical Analysis and Scientific Computing
DU, Cunxiao
ZHOU, Hao
TU, Zhaopeng
JIANG, Jing
Revisiting the Markov Property for machine translation
description In this paper, we re-examine the Markov property in the context of neural machine translation. We design a Markov Autoregressive Transformer (MAT) and undertake a comprehensive assessment of its performance across four WMT benchmarks. Our findings indicate that MAT with an order larger than 4 can generate translations with quality on par with that of conventional autoregressive transformers. In addition, counter-intuitively, we also find that the advantages of utilizing a higher-order MAT do not specifically contribute to the translation of longer sentences.
format text
author DU, Cunxiao
ZHOU, Hao
TU, Zhaopeng
JIANG, Jing
author_facet DU, Cunxiao
ZHOU, Hao
TU, Zhaopeng
JIANG, Jing
author_sort DU, Cunxiao
title Revisiting the Markov Property for machine translation
title_short Revisiting the Markov Property for machine translation
title_full Revisiting the Markov Property for machine translation
title_fullStr Revisiting the Markov Property for machine translation
title_full_unstemmed Revisiting the Markov Property for machine translation
title_sort revisiting the markov property for machine translation
publisher Institutional Knowledge at Singapore Management University
publishDate 2024
url https://ink.library.smu.edu.sg/sis_research/8725
https://ink.library.smu.edu.sg/context/sis_research/article/9728/viewcontent/2024.findings_eacl.40_pvoa.pdf
_version_ 1814047494612451328