@inproceedings{haturusinghe-etal-2025-subasa,
title = "Subasa - Adapting Language Models for Low-resourced Offensive Language Detection in {S}inhala",
author = "Haturusinghe, Shanilka and
Weerasooriya, Tharindu Cyril and
Zampieri, Marcos and
Homan, Christopher M. and
Liyanage, S.R.",
editor = "Ebrahimi, Abteen and
Haider, Samar and
Liu, Emmy and
Haider, Sammar and
Leonor Pacheco, Maria and
Wein, Shira",
booktitle = "Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 4: Student Research Workshop)",
month = apr,
year = "2025",
address = "Albuquerque, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.naacl-srw.26/",
doi = "10.18653/v1/2025.naacl-srw.26",
pages = "260--270",
ISBN = "979-8-89176-192-6",
abstract = "Accurate detection of offensive language is essential for a number of applications related to social media safety. There is a sharp contrast in performance in this task between low and high-resource languages. In this paper, we adapt fine-tuning strategies that have not been previously explored for Sinhala in the downstream task of offensive language detection. Using this approach, we introduce four models: ``Subasa-XLM-R'', which incorporates an intermediate Pre-Finetuning step using Masked Rationale Prediction. Two variants of ``Subasa-Llama'' and ``Subasa-Mistral'', are fine-tuned versions of Llama (3.2) and Mistral (v0.3), respectively, with a task-specific strategy. We evaluate our models on the SOLD benchmark dataset for Sinhala offensive language detection. All our models outperform existing baselines. Subasa-XLM-R achieves the highest Macro F1 score (0.84) surpassing state-of-the-art large language models like GPT-4o when evaluated on the same SOLD benchmark dataset under zero-shot settings. The models and code are publicly available."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="haturusinghe-etal-2025-subasa">
<titleInfo>
<title>Subasa - Adapting Language Models for Low-resourced Offensive Language Detection in Sinhala</title>
</titleInfo>
<name type="personal">
<namePart type="given">Shanilka</namePart>
<namePart type="family">Haturusinghe</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tharindu</namePart>
<namePart type="given">Cyril</namePart>
<namePart type="family">Weerasooriya</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Marcos</namePart>
<namePart type="family">Zampieri</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Christopher</namePart>
<namePart type="given">M</namePart>
<namePart type="family">Homan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">S</namePart>
<namePart type="given">R</namePart>
<namePart type="family">Liyanage</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-04</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 4: Student Research Workshop)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Abteen</namePart>
<namePart type="family">Ebrahimi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Samar</namePart>
<namePart type="family">Haider</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Emmy</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sammar</namePart>
<namePart type="family">Haider</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Maria</namePart>
<namePart type="family">Leonor Pacheco</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shira</namePart>
<namePart type="family">Wein</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Albuquerque, USA</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-192-6</identifier>
</relatedItem>
<abstract>Accurate detection of offensive language is essential for a number of applications related to social media safety. There is a sharp contrast in performance in this task between low and high-resource languages. In this paper, we adapt fine-tuning strategies that have not been previously explored for Sinhala in the downstream task of offensive language detection. Using this approach, we introduce four models: “Subasa-XLM-R”, which incorporates an intermediate Pre-Finetuning step using Masked Rationale Prediction. Two variants of “Subasa-Llama” and “Subasa-Mistral”, are fine-tuned versions of Llama (3.2) and Mistral (v0.3), respectively, with a task-specific strategy. We evaluate our models on the SOLD benchmark dataset for Sinhala offensive language detection. All our models outperform existing baselines. Subasa-XLM-R achieves the highest Macro F1 score (0.84) surpassing state-of-the-art large language models like GPT-4o when evaluated on the same SOLD benchmark dataset under zero-shot settings. The models and code are publicly available.</abstract>
<identifier type="citekey">haturusinghe-etal-2025-subasa</identifier>
<identifier type="doi">10.18653/v1/2025.naacl-srw.26</identifier>
<location>
<url>https://aclanthology.org/2025.naacl-srw.26/</url>
</location>
<part>
<date>2025-04</date>
<extent unit="page">
<start>260</start>
<end>270</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Subasa - Adapting Language Models for Low-resourced Offensive Language Detection in Sinhala
%A Haturusinghe, Shanilka
%A Weerasooriya, Tharindu Cyril
%A Zampieri, Marcos
%A Homan, Christopher M.
%A Liyanage, S. R.
%Y Ebrahimi, Abteen
%Y Haider, Samar
%Y Liu, Emmy
%Y Haider, Sammar
%Y Leonor Pacheco, Maria
%Y Wein, Shira
%S Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 4: Student Research Workshop)
%D 2025
%8 April
%I Association for Computational Linguistics
%C Albuquerque, USA
%@ 979-8-89176-192-6
%F haturusinghe-etal-2025-subasa
%X Accurate detection of offensive language is essential for a number of applications related to social media safety. There is a sharp contrast in performance in this task between low and high-resource languages. In this paper, we adapt fine-tuning strategies that have not been previously explored for Sinhala in the downstream task of offensive language detection. Using this approach, we introduce four models: “Subasa-XLM-R”, which incorporates an intermediate Pre-Finetuning step using Masked Rationale Prediction. Two variants of “Subasa-Llama” and “Subasa-Mistral”, are fine-tuned versions of Llama (3.2) and Mistral (v0.3), respectively, with a task-specific strategy. We evaluate our models on the SOLD benchmark dataset for Sinhala offensive language detection. All our models outperform existing baselines. Subasa-XLM-R achieves the highest Macro F1 score (0.84) surpassing state-of-the-art large language models like GPT-4o when evaluated on the same SOLD benchmark dataset under zero-shot settings. The models and code are publicly available.
%R 10.18653/v1/2025.naacl-srw.26
%U https://aclanthology.org/2025.naacl-srw.26/
%U https://doi.org/10.18653/v1/2025.naacl-srw.26
%P 260-270
Markdown (Informal)
[Subasa - Adapting Language Models for Low-resourced Offensive Language Detection in Sinhala](https://aclanthology.org/2025.naacl-srw.26/) (Haturusinghe et al., NAACL 2025)
ACL
- Shanilka Haturusinghe, Tharindu Cyril Weerasooriya, Marcos Zampieri, Christopher M. Homan, and S.R. Liyanage. 2025. Subasa - Adapting Language Models for Low-resourced Offensive Language Detection in Sinhala. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 4: Student Research Workshop), pages 260–270, Albuquerque, USA. Association for Computational Linguistics.