@inproceedings{mutal-etal-2025-factors,
title = "Factors Affecting Translation Quality in In-context Learning for Multilingual Medical Domain",
author = "Mutal, Jonathan and
Rubino, Raphael and
Bouillon, Pierrette",
editor = "Haddow, Barry and
Kocmi, Tom and
Koehn, Philipp and
Monz, Christof",
booktitle = "Proceedings of the Tenth Conference on Machine Translation",
month = nov,
year = "2025",
address = "Suzhou, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.wmt-1.9/",
pages = "161--179",
ISBN = "979-8-89176-341-8",
abstract = "Multilingual machine translation in the medical domain presents critical challenges due to limited parallel data, domain-specific terminology, and the high stakes associated with translation accuracy. In this paper, we explore the potential of in-context learning (ICL) with general-purpose large language models (LLMs) as an alternative to fine-tuning. Focusing on the medical domain and low-resource languages, we evaluate an instruction-tuned LLM on a translation task across 16 languages. We address four research questions centered on prompt design, examining the impact of the number of examples, the domain and register of examples, and the example selection strategy. Our results show that prompting with one to three examples from the same register and domain as the test input leads to the largest improvements in translation quality, as measured by automatic metrics, while translation quality gains plateau with an increased number of examples. Furthermore, we find that example selection methods - lexical and embedding based - do not yield significant benefits over random selection if the register of selected examples does not match that of the test input."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="mutal-etal-2025-factors">
<titleInfo>
<title>Factors Affecting Translation Quality in In-context Learning for Multilingual Medical Domain</title>
</titleInfo>
<name type="personal">
<namePart type="given">Jonathan</namePart>
<namePart type="family">Mutal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Raphael</namePart>
<namePart type="family">Rubino</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Pierrette</namePart>
<namePart type="family">Bouillon</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the Tenth Conference on Machine Translation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Barry</namePart>
<namePart type="family">Haddow</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tom</namePart>
<namePart type="family">Kocmi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Philipp</namePart>
<namePart type="family">Koehn</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Christof</namePart>
<namePart type="family">Monz</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Suzhou, China</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-341-8</identifier>
</relatedItem>
<abstract>Multilingual machine translation in the medical domain presents critical challenges due to limited parallel data, domain-specific terminology, and the high stakes associated with translation accuracy. In this paper, we explore the potential of in-context learning (ICL) with general-purpose large language models (LLMs) as an alternative to fine-tuning. Focusing on the medical domain and low-resource languages, we evaluate an instruction-tuned LLM on a translation task across 16 languages. We address four research questions centered on prompt design, examining the impact of the number of examples, the domain and register of examples, and the example selection strategy. Our results show that prompting with one to three examples from the same register and domain as the test input leads to the largest improvements in translation quality, as measured by automatic metrics, while translation quality gains plateau with an increased number of examples. Furthermore, we find that example selection methods - lexical and embedding based - do not yield significant benefits over random selection if the register of selected examples does not match that of the test input.</abstract>
<identifier type="citekey">mutal-etal-2025-factors</identifier>
<location>
<url>https://aclanthology.org/2025.wmt-1.9/</url>
</location>
<part>
<date>2025-11</date>
<extent unit="page">
<start>161</start>
<end>179</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Factors Affecting Translation Quality in In-context Learning for Multilingual Medical Domain
%A Mutal, Jonathan
%A Rubino, Raphael
%A Bouillon, Pierrette
%Y Haddow, Barry
%Y Kocmi, Tom
%Y Koehn, Philipp
%Y Monz, Christof
%S Proceedings of the Tenth Conference on Machine Translation
%D 2025
%8 November
%I Association for Computational Linguistics
%C Suzhou, China
%@ 979-8-89176-341-8
%F mutal-etal-2025-factors
%X Multilingual machine translation in the medical domain presents critical challenges due to limited parallel data, domain-specific terminology, and the high stakes associated with translation accuracy. In this paper, we explore the potential of in-context learning (ICL) with general-purpose large language models (LLMs) as an alternative to fine-tuning. Focusing on the medical domain and low-resource languages, we evaluate an instruction-tuned LLM on a translation task across 16 languages. We address four research questions centered on prompt design, examining the impact of the number of examples, the domain and register of examples, and the example selection strategy. Our results show that prompting with one to three examples from the same register and domain as the test input leads to the largest improvements in translation quality, as measured by automatic metrics, while translation quality gains plateau with an increased number of examples. Furthermore, we find that example selection methods - lexical and embedding based - do not yield significant benefits over random selection if the register of selected examples does not match that of the test input.
%U https://aclanthology.org/2025.wmt-1.9/
%P 161-179
Markdown (Informal)
[Factors Affecting Translation Quality in In-context Learning for Multilingual Medical Domain](https://aclanthology.org/2025.wmt-1.9/) (Mutal et al., WMT 2025)
ACL