No Cover Image

Conference Paper/Proceeding/Abstract 301 views 12 downloads

An Exploration-Analysis-Disambiguation Reasoning Framework for Word Sense Disambiguation with Low-Parameter LLMs

Deshan Sumanathilaka Orcid Logo, Nicholas Micallef Orcid Logo, Julian Hough Orcid Logo

Proceedings of the Fifteenth Language Resources and Evaluation Conference (LREC 2026), Pages: 10390 - 10404

Swansea University Authors: Deshan Sumanathilaka Orcid Logo, Nicholas Micallef Orcid Logo, Julian Hough Orcid Logo

  • 71528.VoR.pdf

    PDF | Version of Record

    Licenced under CC-BY-NC-4.0, the Creative Commons Attribution-NonCommercial 4.0 International License.

    Download (389.07KB)

Check full text

DOI (Published version): 10.63317/3oun2fvikwt5

Abstract

Word Sense Disambiguation (WSD) remains a key challenge in Natural Language Processing (NLP), especially when dealing with rare or domain-specific senses that are often misinterpreted. While modern high-parameter Large Language Models (LLMs) such as GPT-4-Turbo have shown state-of-the-art WSD perfor...

Full description

Published in: Proceedings of the Fifteenth Language Resources and Evaluation Conference (LREC 2026)
ISBN: 978-2-493814-49-4
ISSN: 2522-2686
Published: European Language Resources Association (ELRA) 2026
Online Access: Check full text

URI: https://cronfa.swan.ac.uk/Record/cronfa71528
first_indexed 2026-03-03T14:55:09Z
last_indexed 2026-05-16T05:21:41Z
id cronfa71528
recordtype SURis
fullrecord <?xml version="1.0"?><rfc1807><datestamp>2026-05-15T14:36:51.7066627</datestamp><bib-version>v2</bib-version><id>71528</id><entry>2026-03-03</entry><title>An Exploration-Analysis-Disambiguation Reasoning Framework for Word Sense Disambiguation with Low-Parameter LLMs</title><swanseaauthors><author><sid>2fe44f0c1e7d845dc21bb6b00d5b2085</sid><ORCID>0009-0005-8933-6559</ORCID><firstname>Deshan</firstname><surname>Sumanathilaka</surname><name>Deshan Sumanathilaka</name><active>true</active><ethesisStudent>false</ethesisStudent></author><author><sid>1cc4c84582d665b7ee08fb16f5454671</sid><ORCID>0000-0002-2683-8042</ORCID><firstname>Nicholas</firstname><surname>Micallef</surname><name>Nicholas Micallef</name><active>true</active><ethesisStudent>false</ethesisStudent></author><author><sid>082d773ae261d2bbf49434dd2608ab40</sid><ORCID>0000-0002-4345-6759</ORCID><firstname>Julian</firstname><surname>Hough</surname><name>Julian Hough</name><active>true</active><ethesisStudent>false</ethesisStudent></author></swanseaauthors><date>2026-03-03</date><deptcode>MACS</deptcode><abstract>Word Sense Disambiguation (WSD) remains a key challenge in Natural Language Processing (NLP), especially when dealing with rare or domain-specific senses that are often misinterpreted. While modern high-parameter Large Language Models (LLMs) such as GPT-4-Turbo have shown state-of-the-art WSD performance, their computational and energy demands limit scalability. This study investigates whether low-parameter LLMs (&lt;4B parameters) can achieve comparable results through fine-tuning strategies that emphasize reasoning-driven sense identification. Using the FEWS dataset augmented with semi-automated, rationale-rich annotations, we fine-tune eight small-scale open-source LLMs (e.g. Gemma and Qwen). Our results reveal that Chain-of-Thought (CoT)-based reasoning combined with neighbour-word analysis achieves performance comparable to GPT-4-Turbo in zero-shot settings. Importantly, Gemma-3-4B and Qwen-3-4B models consistently outperform all medium-parameter baselines and state-of-the-art models on FEWS, with robust generalization to unseen senses. Furthermore, evaluation on the unseen "Fool Me If You Can&#x201D; dataset confirms strong cross-domain adaptability without task-specific fine-tuning. This work demonstrates that with carefully crafted reasoning-centric fine-tuning, low-parameter LLMs can deliver accurate WSD while substantially reducing computational and energy demands.</abstract><type>Conference Paper/Proceeding/Abstract</type><journal>Proceedings of the Fifteenth Language Resources and Evaluation Conference (LREC 2026)</journal><volume/><journalNumber/><paginationStart>10390</paginationStart><paginationEnd>10404</paginationEnd><publisher>European Language Resources Association (ELRA)</publisher><placeOfPublication/><isbnPrint>978-2-493814-49-4</isbnPrint><isbnElectronic/><issnPrint>2522-2686</issnPrint><issnElectronic/><keywords>Word Sense Disambiguation, Low-parameter LLMs, Reasoning-driven Fine-tuning</keywords><publishedDay>11</publishedDay><publishedMonth>5</publishedMonth><publishedYear>2026</publishedYear><publishedDate>2026-05-11</publishedDate><doi>10.63317/3oun2fvikwt5</doi><url/><notes/><college>COLLEGE NANME</college><department>Mathematics and Computer Science School</department><CollegeCode>COLLEGE CODE</CollegeCode><DepartmentCode>MACS</DepartmentCode><institution>Swansea University</institution><apcterm>Other</apcterm><funders>We acknowledge the support of the Super computing Wales project, which is part-funded by the European Regional Development Fund (ERDF) via Welsh Government. Hough&#x2019;s work is supported by the EPSRC grant EP/X009343/1 &#x2018;FLUIDITY&#x2019;.</funders><projectreference/><lastEdited>2026-05-15T14:36:51.7066627</lastEdited><Created>2026-03-03T14:47:58.7796742</Created><path><level id="1">Faculty of Science and Engineering</level><level id="2">School of Mathematics and Computer Science - Computer Science</level></path><authors><author><firstname>Deshan</firstname><surname>Sumanathilaka</surname><orcid>0009-0005-8933-6559</orcid><order>1</order></author><author><firstname>Nicholas</firstname><surname>Micallef</surname><orcid>0000-0002-2683-8042</orcid><order>2</order></author><author><firstname>Julian</firstname><surname>Hough</surname><orcid>0000-0002-4345-6759</orcid><order>3</order></author></authors><documents><document><filename>71528__36750__f03d0241772648e586719beb1fceed7f.pdf</filename><originalFilename>71528.VoR.pdf</originalFilename><uploaded>2026-05-15T14:24:16.0766958</uploaded><type>Output</type><contentLength>398406</contentLength><contentType>application/pdf</contentType><version>Version of Record</version><cronfaStatus>true</cronfaStatus><documentNotes>Licenced under CC-BY-NC-4.0, the Creative Commons Attribution-NonCommercial 4.0 International License.</documentNotes><copyrightCorrect>true</copyrightCorrect><language>eng</language><licence>https://creativecommons.org/licenses/by-nc/4.0/</licence></document></documents><OutputDurs/></rfc1807>
spelling 2026-05-15T14:36:51.7066627 v2 71528 2026-03-03 An Exploration-Analysis-Disambiguation Reasoning Framework for Word Sense Disambiguation with Low-Parameter LLMs 2fe44f0c1e7d845dc21bb6b00d5b2085 0009-0005-8933-6559 Deshan Sumanathilaka Deshan Sumanathilaka true false 1cc4c84582d665b7ee08fb16f5454671 0000-0002-2683-8042 Nicholas Micallef Nicholas Micallef true false 082d773ae261d2bbf49434dd2608ab40 0000-0002-4345-6759 Julian Hough Julian Hough true false 2026-03-03 MACS Word Sense Disambiguation (WSD) remains a key challenge in Natural Language Processing (NLP), especially when dealing with rare or domain-specific senses that are often misinterpreted. While modern high-parameter Large Language Models (LLMs) such as GPT-4-Turbo have shown state-of-the-art WSD performance, their computational and energy demands limit scalability. This study investigates whether low-parameter LLMs (<4B parameters) can achieve comparable results through fine-tuning strategies that emphasize reasoning-driven sense identification. Using the FEWS dataset augmented with semi-automated, rationale-rich annotations, we fine-tune eight small-scale open-source LLMs (e.g. Gemma and Qwen). Our results reveal that Chain-of-Thought (CoT)-based reasoning combined with neighbour-word analysis achieves performance comparable to GPT-4-Turbo in zero-shot settings. Importantly, Gemma-3-4B and Qwen-3-4B models consistently outperform all medium-parameter baselines and state-of-the-art models on FEWS, with robust generalization to unseen senses. Furthermore, evaluation on the unseen "Fool Me If You Can” dataset confirms strong cross-domain adaptability without task-specific fine-tuning. This work demonstrates that with carefully crafted reasoning-centric fine-tuning, low-parameter LLMs can deliver accurate WSD while substantially reducing computational and energy demands. Conference Paper/Proceeding/Abstract Proceedings of the Fifteenth Language Resources and Evaluation Conference (LREC 2026) 10390 10404 European Language Resources Association (ELRA) 978-2-493814-49-4 2522-2686 Word Sense Disambiguation, Low-parameter LLMs, Reasoning-driven Fine-tuning 11 5 2026 2026-05-11 10.63317/3oun2fvikwt5 COLLEGE NANME Mathematics and Computer Science School COLLEGE CODE MACS Swansea University Other We acknowledge the support of the Super computing Wales project, which is part-funded by the European Regional Development Fund (ERDF) via Welsh Government. Hough’s work is supported by the EPSRC grant EP/X009343/1 ‘FLUIDITY’. 2026-05-15T14:36:51.7066627 2026-03-03T14:47:58.7796742 Faculty of Science and Engineering School of Mathematics and Computer Science - Computer Science Deshan Sumanathilaka 0009-0005-8933-6559 1 Nicholas Micallef 0000-0002-2683-8042 2 Julian Hough 0000-0002-4345-6759 3 71528__36750__f03d0241772648e586719beb1fceed7f.pdf 71528.VoR.pdf 2026-05-15T14:24:16.0766958 Output 398406 application/pdf Version of Record true Licenced under CC-BY-NC-4.0, the Creative Commons Attribution-NonCommercial 4.0 International License. true eng https://creativecommons.org/licenses/by-nc/4.0/
title An Exploration-Analysis-Disambiguation Reasoning Framework for Word Sense Disambiguation with Low-Parameter LLMs
spellingShingle An Exploration-Analysis-Disambiguation Reasoning Framework for Word Sense Disambiguation with Low-Parameter LLMs
Deshan Sumanathilaka
Nicholas Micallef
Julian Hough
title_short An Exploration-Analysis-Disambiguation Reasoning Framework for Word Sense Disambiguation with Low-Parameter LLMs
title_full An Exploration-Analysis-Disambiguation Reasoning Framework for Word Sense Disambiguation with Low-Parameter LLMs
title_fullStr An Exploration-Analysis-Disambiguation Reasoning Framework for Word Sense Disambiguation with Low-Parameter LLMs
title_full_unstemmed An Exploration-Analysis-Disambiguation Reasoning Framework for Word Sense Disambiguation with Low-Parameter LLMs
title_sort An Exploration-Analysis-Disambiguation Reasoning Framework for Word Sense Disambiguation with Low-Parameter LLMs
author_id_str_mv 2fe44f0c1e7d845dc21bb6b00d5b2085
1cc4c84582d665b7ee08fb16f5454671
082d773ae261d2bbf49434dd2608ab40
author_id_fullname_str_mv 2fe44f0c1e7d845dc21bb6b00d5b2085_***_Deshan Sumanathilaka
1cc4c84582d665b7ee08fb16f5454671_***_Nicholas Micallef
082d773ae261d2bbf49434dd2608ab40_***_Julian Hough
author Deshan Sumanathilaka
Nicholas Micallef
Julian Hough
author2 Deshan Sumanathilaka
Nicholas Micallef
Julian Hough
format Conference Paper/Proceeding/Abstract
container_title Proceedings of the Fifteenth Language Resources and Evaluation Conference (LREC 2026)
container_start_page 10390
publishDate 2026
institution Swansea University
isbn 978-2-493814-49-4
issn 2522-2686
doi_str_mv 10.63317/3oun2fvikwt5
publisher European Language Resources Association (ELRA)
college_str Faculty of Science and Engineering
hierarchytype
hierarchy_top_id facultyofscienceandengineering
hierarchy_top_title Faculty of Science and Engineering
hierarchy_parent_id facultyofscienceandengineering
hierarchy_parent_title Faculty of Science and Engineering
department_str School of Mathematics and Computer Science - Computer Science{{{_:::_}}}Faculty of Science and Engineering{{{_:::_}}}School of Mathematics and Computer Science - Computer Science
document_store_str 1
active_str 0
description Word Sense Disambiguation (WSD) remains a key challenge in Natural Language Processing (NLP), especially when dealing with rare or domain-specific senses that are often misinterpreted. While modern high-parameter Large Language Models (LLMs) such as GPT-4-Turbo have shown state-of-the-art WSD performance, their computational and energy demands limit scalability. This study investigates whether low-parameter LLMs (<4B parameters) can achieve comparable results through fine-tuning strategies that emphasize reasoning-driven sense identification. Using the FEWS dataset augmented with semi-automated, rationale-rich annotations, we fine-tune eight small-scale open-source LLMs (e.g. Gemma and Qwen). Our results reveal that Chain-of-Thought (CoT)-based reasoning combined with neighbour-word analysis achieves performance comparable to GPT-4-Turbo in zero-shot settings. Importantly, Gemma-3-4B and Qwen-3-4B models consistently outperform all medium-parameter baselines and state-of-the-art models on FEWS, with robust generalization to unseen senses. Furthermore, evaluation on the unseen "Fool Me If You Can” dataset confirms strong cross-domain adaptability without task-specific fine-tuning. This work demonstrates that with carefully crafted reasoning-centric fine-tuning, low-parameter LLMs can deliver accurate WSD while substantially reducing computational and energy demands.
published_date 2026-05-11T17:19:12Z
_version_ 1866630905112559616
score 11.106612

Similar Items