@inproceedings{kermani-etal-2025-finetuning,
title = "Finetuning Pre-trained Language Models for Bidirectional Sign Language Gloss to Text Translation",
author = "Kermani, Arshia and
Irani, Habib and
Metsis, Vangelis",
editor = "Hasanuzzaman, Mohammed and
Quiroga, Facundo Manuel and
Modi, Ashutosh and
Kamila, Sabyasachi and
Artiaga, Keren and
Joshi, Abhinav and
Singh, Sanjeet",
booktitle = "Proceedings of the Workshop on Sign Language Processing (WSLP)",
month = dec,
year = "2025",
address = "IIT Bombay, Mumbai, India (Co-located with IJCNLP{--}AACL 2025)",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.wslp-main.11/",
pages = "73--81",
ISBN = "979-8-89176-304-3",
abstract = "Sign Language Translation (SLT) is a crucial technology for fostering communication accessibility for the Deaf and Hard-of-Hearing (DHH) community. A dominant approach in SLT involves a two-stage pipeline: first, transcribing video to sign language glosses, and then translating these glosses into natural text. This second stage, gloss-to-text translation, is a challenging, low-resource machine translation task due to data scarcity and significant syntactic divergence. While prior work has often relied on training translation models from scratch, we show that fine-tuning large, pre-trained language models (PLMs) offers a more effective and data-efficient paradigm. In this work, we conduct a comprehensive bidirectional evaluation of several PLMs (T5, Flan-T5, mBART, and Llama) on this task. We use a collection of popular SLT datasets (RWTH-PHOENIX-14T, SIGNUM, and ASLG-PC12) and evaluate performance using standard machine translation metrics. Our results show that fine-tuned PLMs consistently and significantly outperform Transformer models trained from scratch, establishing new state-of-the-art results. Crucially, our bidirectional analysis reveals a significant performance gap, with Text-to-Gloss translation posing a greater challenge than Gloss-to-Text. We conclude that leveraging the linguistic knowledge of pre-trained models is a superior strategy for gloss translation and provides a more practical foundation for building robust, real-world SLT systems."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="kermani-etal-2025-finetuning">
<titleInfo>
<title>Finetuning Pre-trained Language Models for Bidirectional Sign Language Gloss to Text Translation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Arshia</namePart>
<namePart type="family">Kermani</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Habib</namePart>
<namePart type="family">Irani</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vangelis</namePart>
<namePart type="family">Metsis</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-12</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the Workshop on Sign Language Processing (WSLP)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Mohammed</namePart>
<namePart type="family">Hasanuzzaman</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Facundo</namePart>
<namePart type="given">Manuel</namePart>
<namePart type="family">Quiroga</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ashutosh</namePart>
<namePart type="family">Modi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sabyasachi</namePart>
<namePart type="family">Kamila</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Keren</namePart>
<namePart type="family">Artiaga</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Abhinav</namePart>
<namePart type="family">Joshi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sanjeet</namePart>
<namePart type="family">Singh</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">IIT Bombay, Mumbai, India (Co-located with IJCNLP–AACL 2025)</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-304-3</identifier>
</relatedItem>
<abstract>Sign Language Translation (SLT) is a crucial technology for fostering communication accessibility for the Deaf and Hard-of-Hearing (DHH) community. A dominant approach in SLT involves a two-stage pipeline: first, transcribing video to sign language glosses, and then translating these glosses into natural text. This second stage, gloss-to-text translation, is a challenging, low-resource machine translation task due to data scarcity and significant syntactic divergence. While prior work has often relied on training translation models from scratch, we show that fine-tuning large, pre-trained language models (PLMs) offers a more effective and data-efficient paradigm. In this work, we conduct a comprehensive bidirectional evaluation of several PLMs (T5, Flan-T5, mBART, and Llama) on this task. We use a collection of popular SLT datasets (RWTH-PHOENIX-14T, SIGNUM, and ASLG-PC12) and evaluate performance using standard machine translation metrics. Our results show that fine-tuned PLMs consistently and significantly outperform Transformer models trained from scratch, establishing new state-of-the-art results. Crucially, our bidirectional analysis reveals a significant performance gap, with Text-to-Gloss translation posing a greater challenge than Gloss-to-Text. We conclude that leveraging the linguistic knowledge of pre-trained models is a superior strategy for gloss translation and provides a more practical foundation for building robust, real-world SLT systems.</abstract>
<identifier type="citekey">kermani-etal-2025-finetuning</identifier>
<location>
<url>https://aclanthology.org/2025.wslp-main.11/</url>
</location>
<part>
<date>2025-12</date>
<extent unit="page">
<start>73</start>
<end>81</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Finetuning Pre-trained Language Models for Bidirectional Sign Language Gloss to Text Translation
%A Kermani, Arshia
%A Irani, Habib
%A Metsis, Vangelis
%Y Hasanuzzaman, Mohammed
%Y Quiroga, Facundo Manuel
%Y Modi, Ashutosh
%Y Kamila, Sabyasachi
%Y Artiaga, Keren
%Y Joshi, Abhinav
%Y Singh, Sanjeet
%S Proceedings of the Workshop on Sign Language Processing (WSLP)
%D 2025
%8 December
%I Association for Computational Linguistics
%C IIT Bombay, Mumbai, India (Co-located with IJCNLP–AACL 2025)
%@ 979-8-89176-304-3
%F kermani-etal-2025-finetuning
%X Sign Language Translation (SLT) is a crucial technology for fostering communication accessibility for the Deaf and Hard-of-Hearing (DHH) community. A dominant approach in SLT involves a two-stage pipeline: first, transcribing video to sign language glosses, and then translating these glosses into natural text. This second stage, gloss-to-text translation, is a challenging, low-resource machine translation task due to data scarcity and significant syntactic divergence. While prior work has often relied on training translation models from scratch, we show that fine-tuning large, pre-trained language models (PLMs) offers a more effective and data-efficient paradigm. In this work, we conduct a comprehensive bidirectional evaluation of several PLMs (T5, Flan-T5, mBART, and Llama) on this task. We use a collection of popular SLT datasets (RWTH-PHOENIX-14T, SIGNUM, and ASLG-PC12) and evaluate performance using standard machine translation metrics. Our results show that fine-tuned PLMs consistently and significantly outperform Transformer models trained from scratch, establishing new state-of-the-art results. Crucially, our bidirectional analysis reveals a significant performance gap, with Text-to-Gloss translation posing a greater challenge than Gloss-to-Text. We conclude that leveraging the linguistic knowledge of pre-trained models is a superior strategy for gloss translation and provides a more practical foundation for building robust, real-world SLT systems.
%U https://aclanthology.org/2025.wslp-main.11/
%P 73-81
Markdown (Informal)
[Finetuning Pre-trained Language Models for Bidirectional Sign Language Gloss to Text Translation](https://aclanthology.org/2025.wslp-main.11/) (Kermani et al., WSLP 2025)
ACL