@inproceedings{aurelian-tjiaranata-etal-2025-buinus,
title = "{BUINUS} at {IWSLT}: Evaluating the Impact of Data Augmentation and {QL}o{RA}-based Fine-Tuning for {M}altese to {E}nglish Speech Translation",
author = "Aurelian Tjiaranata, Filbert and
Alexandra Putra, Vallerie and
Presma Yulianrifat, Eryawan and
Akmal Hanif, Ikhlasul",
editor = "Salesky, Elizabeth and
Federico, Marcello and
Anastasopoulos, Antonis",
booktitle = "Proceedings of the 22nd International Conference on Spoken Language Translation (IWSLT 2025)",
month = jul,
year = "2025",
address = "Vienna, Austria (in-person and online)",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.iwslt-1.26/",
doi = "10.18653/v1/2025.iwslt-1.26",
pages = "269--273",
ISBN = "979-8-89176-272-5",
abstract = "This paper investigates approaches for the IWSLT low-resource track, Track 1 (speech-to-text translation) for the Maltese language, focusing on data augmentation and large pre-trained models. Our system combines Whisper for transcription and NLLB for translation, with experiments concentrated mainly on the translation stage. We observe that data augmentation leads to only marginal improvements, primarily for the smaller 600M model, with gains up to 0.0026 COMET points. These gains do not extend to larger models like the 3.3B NLLB, and the overall impact appears somewhat inconsistent. In contrast, fine-tuning larger models using QLoRA outperforms full fine-tuning of smaller models. Moreover, multi-stage fine-tuning consistently improves task-specific performance across all model sizes."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="aurelian-tjiaranata-etal-2025-buinus">
<titleInfo>
<title>BUINUS at IWSLT: Evaluating the Impact of Data Augmentation and QLoRA-based Fine-Tuning for Maltese to English Speech Translation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Filbert</namePart>
<namePart type="family">Aurelian Tjiaranata</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vallerie</namePart>
<namePart type="family">Alexandra Putra</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Eryawan</namePart>
<namePart type="family">Presma Yulianrifat</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ikhlasul</namePart>
<namePart type="family">Akmal Hanif</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 22nd International Conference on Spoken Language Translation (IWSLT 2025)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Elizabeth</namePart>
<namePart type="family">Salesky</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Marcello</namePart>
<namePart type="family">Federico</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Antonis</namePart>
<namePart type="family">Anastasopoulos</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Vienna, Austria (in-person and online)</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-272-5</identifier>
</relatedItem>
<abstract>This paper investigates approaches for the IWSLT low-resource track, Track 1 (speech-to-text translation) for the Maltese language, focusing on data augmentation and large pre-trained models. Our system combines Whisper for transcription and NLLB for translation, with experiments concentrated mainly on the translation stage. We observe that data augmentation leads to only marginal improvements, primarily for the smaller 600M model, with gains up to 0.0026 COMET points. These gains do not extend to larger models like the 3.3B NLLB, and the overall impact appears somewhat inconsistent. In contrast, fine-tuning larger models using QLoRA outperforms full fine-tuning of smaller models. Moreover, multi-stage fine-tuning consistently improves task-specific performance across all model sizes.</abstract>
<identifier type="citekey">aurelian-tjiaranata-etal-2025-buinus</identifier>
<identifier type="doi">10.18653/v1/2025.iwslt-1.26</identifier>
<location>
<url>https://aclanthology.org/2025.iwslt-1.26/</url>
</location>
<part>
<date>2025-07</date>
<extent unit="page">
<start>269</start>
<end>273</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T BUINUS at IWSLT: Evaluating the Impact of Data Augmentation and QLoRA-based Fine-Tuning for Maltese to English Speech Translation
%A Aurelian Tjiaranata, Filbert
%A Alexandra Putra, Vallerie
%A Presma Yulianrifat, Eryawan
%A Akmal Hanif, Ikhlasul
%Y Salesky, Elizabeth
%Y Federico, Marcello
%Y Anastasopoulos, Antonis
%S Proceedings of the 22nd International Conference on Spoken Language Translation (IWSLT 2025)
%D 2025
%8 July
%I Association for Computational Linguistics
%C Vienna, Austria (in-person and online)
%@ 979-8-89176-272-5
%F aurelian-tjiaranata-etal-2025-buinus
%X This paper investigates approaches for the IWSLT low-resource track, Track 1 (speech-to-text translation) for the Maltese language, focusing on data augmentation and large pre-trained models. Our system combines Whisper for transcription and NLLB for translation, with experiments concentrated mainly on the translation stage. We observe that data augmentation leads to only marginal improvements, primarily for the smaller 600M model, with gains up to 0.0026 COMET points. These gains do not extend to larger models like the 3.3B NLLB, and the overall impact appears somewhat inconsistent. In contrast, fine-tuning larger models using QLoRA outperforms full fine-tuning of smaller models. Moreover, multi-stage fine-tuning consistently improves task-specific performance across all model sizes.
%R 10.18653/v1/2025.iwslt-1.26
%U https://aclanthology.org/2025.iwslt-1.26/
%U https://doi.org/10.18653/v1/2025.iwslt-1.26
%P 269-273
Markdown (Informal)
[BUINUS at IWSLT: Evaluating the Impact of Data Augmentation and QLoRA-based Fine-Tuning for Maltese to English Speech Translation](https://aclanthology.org/2025.iwslt-1.26/) (Aurelian Tjiaranata et al., IWSLT 2025)
ACL