@inproceedings{jeon-etal-2026-phonemizing,
title = "Phonemizing User-Generated Text: A Benchmark, Taxonomy, and Compositional Approach",
author = "Jeon, MinJu and
Park, Younghan and
Park, Han Sung and
Kim, Jong-Hwan and
Kim, Dong-Jin and
Lee, Hoyeon",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.1020/",
pages = "21564--21581",
ISBN = "979-8-89176-462-0",
abstract = "Text-to-speech systems increasingly process user-generated text (UGT) such as ppl and imo, whose pronunciation must be inferred from the canonical rather than surface form. We introduce UGTPHON, the first grapheme-to-phoneme (G2P) benchmark for UGT in English, Vietnamese, and Korean, with an inference-grounded taxonomy. Existing G2P models and frontier LLMs share a systematic failure pattern, with the canonical-to-non-canonical gap reaching up to 66.8 PER points. We propose a retrieval-augmented G2P approach that explicitly models canonical-form inference through retrieval and staged decoding. Under 0.5B parameters, our approach shrinks this gap by up to 83{\%} while outperforming most frontier LLMs that are orders of magnitude larger, demonstrating that explicit compositional structure can substitute for scale."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="jeon-etal-2026-phonemizing">
<titleInfo>
<title>Phonemizing User-Generated Text: A Benchmark, Taxonomy, and Compositional Approach</title>
</titleInfo>
<name type="personal">
<namePart type="given">MinJu</namePart>
<namePart type="family">Jeon</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Younghan</namePart>
<namePart type="family">Park</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Han</namePart>
<namePart type="given">Sung</namePart>
<namePart type="family">Park</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jong-Hwan</namePart>
<namePart type="family">Kim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dong-Jin</namePart>
<namePart type="family">Kim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hoyeon</namePart>
<namePart type="family">Lee</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>Findings of the Association for Computational Linguistics: EMNLP 2026</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">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>Text-to-speech systems increasingly process user-generated text (UGT) such as ppl and imo, whose pronunciation must be inferred from the canonical rather than surface form. We introduce UGTPHON, the first grapheme-to-phoneme (G2P) benchmark for UGT in English, Vietnamese, and Korean, with an inference-grounded taxonomy. Existing G2P models and frontier LLMs share a systematic failure pattern, with the canonical-to-non-canonical gap reaching up to 66.8 PER points. We propose a retrieval-augmented G2P approach that explicitly models canonical-form inference through retrieval and staged decoding. Under 0.5B parameters, our approach shrinks this gap by up to 83% while outperforming most frontier LLMs that are orders of magnitude larger, demonstrating that explicit compositional structure can substitute for scale.</abstract>
<identifier type="citekey">jeon-etal-2026-phonemizing</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.1020/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>21564</start>
<end>21581</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Phonemizing User-Generated Text: A Benchmark, Taxonomy, and Compositional Approach
%A Jeon, MinJu
%A Park, Younghan
%A Park, Han Sung
%A Kim, Jong-Hwan
%A Kim, Dong-Jin
%A Lee, Hoyeon
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F jeon-etal-2026-phonemizing
%X Text-to-speech systems increasingly process user-generated text (UGT) such as ppl and imo, whose pronunciation must be inferred from the canonical rather than surface form. We introduce UGTPHON, the first grapheme-to-phoneme (G2P) benchmark for UGT in English, Vietnamese, and Korean, with an inference-grounded taxonomy. Existing G2P models and frontier LLMs share a systematic failure pattern, with the canonical-to-non-canonical gap reaching up to 66.8 PER points. We propose a retrieval-augmented G2P approach that explicitly models canonical-form inference through retrieval and staged decoding. Under 0.5B parameters, our approach shrinks this gap by up to 83% while outperforming most frontier LLMs that are orders of magnitude larger, demonstrating that explicit compositional structure can substitute for scale.
%U https://aclanthology.org/2026.findings-emnlp.1020/
%P 21564-21581
Markdown (Informal)
[Phonemizing User-Generated Text: A Benchmark, Taxonomy, and Compositional Approach](https://aclanthology.org/2026.findings-emnlp.1020/) (Jeon et al., Findings 2026)
ACL
- MinJu Jeon, Younghan Park, Han Sung Park, Jong-Hwan Kim, Dong-Jin Kim, and Hoyeon Lee. 2026. Phonemizing User-Generated Text: A Benchmark, Taxonomy, and Compositional Approach. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 21564–21581, Budapest, Hungary. Association for Computational Linguistics.