@inproceedings{vaflor-roxas-2026-cebidn,
title = "{C}eb{IDN}: A Computational Framework for {C}ebuano Idiom Identification, Disambiguation, and Normalization",
author = "Vaflor, Adrian P. and
Roxas, Robert R.",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.2318/",
pages = "52655--52682",
ISBN = "979-8-89176-459-0",
abstract = "Idioms remain a persistent challenge for natural language understanding due to their non-compositional nature, especially in morphologically rich, low-resource languages like Cebuano. Existing work on idiom processing has explored span extraction, contrastive learning, and normalization, but typically treats these tasks independently or assumes previous steps have already been done. We propose CebIDN, a modular pipeline that unifies idiom identification, disambiguation, and normalization for Cebuano. Central to our framework is a morphology-aware canonical matcher that links extracted spans to an idiom knowledge base, passing structural tags, the idiom used, and retrieved literal hints to the downstream generative model. The system is evaluated in various tagging configurations, incorporating CRF-based sequence labeling and span-level triplet loss, along with few-shot and fine-tuned normalization strategies across both tagged and tag-free configurations. The optimal configuration (using a 3-tag BIO+CRF tagger with a tag-free normalizer) significantly outperforms a fine-tuned end-to-end LLM baseline by improving binary idiom detection (0.899 vs. 0.847 F1) and achieving superior normalization quality (88.87 vs. 86.89 BLEU; 96.40 vs. 95.04 BERTScore; 4.71 vs. 4.46 human-validated LLM evaluation scores), while running 1.7 times faster. These results demonstrate that structured task decoupling improves figurative language understanding in low-resource settings."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="vaflor-roxas-2026-cebidn">
<titleInfo>
<title>CebIDN: A Computational Framework for Cebuano Idiom Identification, Disambiguation, and Normalization</title>
</titleInfo>
<name type="personal">
<namePart type="given">Adrian</namePart>
<namePart type="given">P</namePart>
<namePart type="family">Vaflor</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Robert</namePart>
<namePart type="given">R</namePart>
<namePart type="family">Roxas</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>Idioms remain a persistent challenge for natural language understanding due to their non-compositional nature, especially in morphologically rich, low-resource languages like Cebuano. Existing work on idiom processing has explored span extraction, contrastive learning, and normalization, but typically treats these tasks independently or assumes previous steps have already been done. We propose CebIDN, a modular pipeline that unifies idiom identification, disambiguation, and normalization for Cebuano. Central to our framework is a morphology-aware canonical matcher that links extracted spans to an idiom knowledge base, passing structural tags, the idiom used, and retrieved literal hints to the downstream generative model. The system is evaluated in various tagging configurations, incorporating CRF-based sequence labeling and span-level triplet loss, along with few-shot and fine-tuned normalization strategies across both tagged and tag-free configurations. The optimal configuration (using a 3-tag BIO+CRF tagger with a tag-free normalizer) significantly outperforms a fine-tuned end-to-end LLM baseline by improving binary idiom detection (0.899 vs. 0.847 F1) and achieving superior normalization quality (88.87 vs. 86.89 BLEU; 96.40 vs. 95.04 BERTScore; 4.71 vs. 4.46 human-validated LLM evaluation scores), while running 1.7 times faster. These results demonstrate that structured task decoupling improves figurative language understanding in low-resource settings.</abstract>
<identifier type="citekey">vaflor-roxas-2026-cebidn</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.2318/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>52655</start>
<end>52682</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T CebIDN: A Computational Framework for Cebuano Idiom Identification, Disambiguation, and Normalization
%A Vaflor, Adrian P.
%A Roxas, Robert R.
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F vaflor-roxas-2026-cebidn
%X Idioms remain a persistent challenge for natural language understanding due to their non-compositional nature, especially in morphologically rich, low-resource languages like Cebuano. Existing work on idiom processing has explored span extraction, contrastive learning, and normalization, but typically treats these tasks independently or assumes previous steps have already been done. We propose CebIDN, a modular pipeline that unifies idiom identification, disambiguation, and normalization for Cebuano. Central to our framework is a morphology-aware canonical matcher that links extracted spans to an idiom knowledge base, passing structural tags, the idiom used, and retrieved literal hints to the downstream generative model. The system is evaluated in various tagging configurations, incorporating CRF-based sequence labeling and span-level triplet loss, along with few-shot and fine-tuned normalization strategies across both tagged and tag-free configurations. The optimal configuration (using a 3-tag BIO+CRF tagger with a tag-free normalizer) significantly outperforms a fine-tuned end-to-end LLM baseline by improving binary idiom detection (0.899 vs. 0.847 F1) and achieving superior normalization quality (88.87 vs. 86.89 BLEU; 96.40 vs. 95.04 BERTScore; 4.71 vs. 4.46 human-validated LLM evaluation scores), while running 1.7 times faster. These results demonstrate that structured task decoupling improves figurative language understanding in low-resource settings.
%U https://aclanthology.org/2026.emnlp-main.2318/
%P 52655-52682
Markdown (Informal)
[CebIDN: A Computational Framework for Cebuano Idiom Identification, Disambiguation, and Normalization](https://aclanthology.org/2026.emnlp-main.2318/) (Vaflor & Roxas, EMNLP 2026)
ACL