@inproceedings{rachmat-etal-2026-semantic,
title = "Semantic, Syntactic, Lexical: What Makes {QA} Augmentation Work in Limited Quantity?",
author = "Rachmat, Benedictus Kent and
Gerald, Thomas and
Nakamura, Takuya and
Zhang, Zheng and
Grouin, Cyril",
editor = "Hinrichs, Erhard and
Nivre, Joakim and
Osenova, Petya and
Pustejovsky, James and
Zinn, Claus",
booktitle = "Proceedings of the Workshop on Structured Linguistic Data and Evaluation ({SL}i{DE})",
month = may,
year = "2026",
address = "Palma de Mallorca, Spain",
publisher = "ELRA Language Resources Association (ELRA)",
url = "https://aclanthology.org/2026.slide-1.20/",
doi = "10.63317/2545w4h6dty8",
pages = "224--236",
abstract = "Data augmentation is a common fix in domains where training data is scarce or difficult to collect, such as specialized medical or any other domain specific applications. In question answering (QA), most studies report headline accuracy while saying little about the quality of the synthetic data. Here, quality goes beyond fluent rewording: augmented items must remain faithful to the supporting evidence and preserve the original answerability. We study three augmentation families lexical, syntactic, and semantic edits generated with LLaMA 3.1 70B, and analyze how these edits affect model behavior. To mirror low-resource settings, we focus on subsets of SQuADv2 (general) and PubMedQA (biomedical, domain specific). We report Exact Match (EM)/F1 alongside quality diagnostics, yielding a fuller picture than accuracy alone. Our results show that augmentation behaves differently across domains and scales. In SQuADv2, augmented variants maintain performance on par with baselines, showing that added diversity mostly does not harm model quality, whereas in PubMedQA semantic edits bring improvements under extreme scarcity and support stronger performance as supervision grows."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="rachmat-etal-2026-semantic">
<titleInfo>
<title>Semantic, Syntactic, Lexical: What Makes QA Augmentation Work in Limited Quantity?</title>
</titleInfo>
<name type="personal">
<namePart type="given">Benedictus</namePart>
<namePart type="given">Kent</namePart>
<namePart type="family">Rachmat</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Thomas</namePart>
<namePart type="family">Gerald</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Takuya</namePart>
<namePart type="family">Nakamura</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zheng</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Cyril</namePart>
<namePart type="family">Grouin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-05</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the Workshop on Structured Linguistic Data and Evaluation (SLiDE)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Erhard</namePart>
<namePart type="family">Hinrichs</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Joakim</namePart>
<namePart type="family">Nivre</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Petya</namePart>
<namePart type="family">Osenova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">James</namePart>
<namePart type="family">Pustejovsky</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Claus</namePart>
<namePart type="family">Zinn</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>ELRA Language Resources Association (ELRA)</publisher>
<place>
<placeTerm type="text">Palma de Mallorca, Spain</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Data augmentation is a common fix in domains where training data is scarce or difficult to collect, such as specialized medical or any other domain specific applications. In question answering (QA), most studies report headline accuracy while saying little about the quality of the synthetic data. Here, quality goes beyond fluent rewording: augmented items must remain faithful to the supporting evidence and preserve the original answerability. We study three augmentation families lexical, syntactic, and semantic edits generated with LLaMA 3.1 70B, and analyze how these edits affect model behavior. To mirror low-resource settings, we focus on subsets of SQuADv2 (general) and PubMedQA (biomedical, domain specific). We report Exact Match (EM)/F1 alongside quality diagnostics, yielding a fuller picture than accuracy alone. Our results show that augmentation behaves differently across domains and scales. In SQuADv2, augmented variants maintain performance on par with baselines, showing that added diversity mostly does not harm model quality, whereas in PubMedQA semantic edits bring improvements under extreme scarcity and support stronger performance as supervision grows.</abstract>
<identifier type="citekey">rachmat-etal-2026-semantic</identifier>
<identifier type="doi">10.63317/2545w4h6dty8</identifier>
<location>
<url>https://aclanthology.org/2026.slide-1.20/</url>
</location>
<part>
<date>2026-05</date>
<extent unit="page">
<start>224</start>
<end>236</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Semantic, Syntactic, Lexical: What Makes QA Augmentation Work in Limited Quantity?
%A Rachmat, Benedictus Kent
%A Gerald, Thomas
%A Nakamura, Takuya
%A Zhang, Zheng
%A Grouin, Cyril
%Y Hinrichs, Erhard
%Y Nivre, Joakim
%Y Osenova, Petya
%Y Pustejovsky, James
%Y Zinn, Claus
%S Proceedings of the Workshop on Structured Linguistic Data and Evaluation (SLiDE)
%D 2026
%8 May
%I ELRA Language Resources Association (ELRA)
%C Palma de Mallorca, Spain
%F rachmat-etal-2026-semantic
%X Data augmentation is a common fix in domains where training data is scarce or difficult to collect, such as specialized medical or any other domain specific applications. In question answering (QA), most studies report headline accuracy while saying little about the quality of the synthetic data. Here, quality goes beyond fluent rewording: augmented items must remain faithful to the supporting evidence and preserve the original answerability. We study three augmentation families lexical, syntactic, and semantic edits generated with LLaMA 3.1 70B, and analyze how these edits affect model behavior. To mirror low-resource settings, we focus on subsets of SQuADv2 (general) and PubMedQA (biomedical, domain specific). We report Exact Match (EM)/F1 alongside quality diagnostics, yielding a fuller picture than accuracy alone. Our results show that augmentation behaves differently across domains and scales. In SQuADv2, augmented variants maintain performance on par with baselines, showing that added diversity mostly does not harm model quality, whereas in PubMedQA semantic edits bring improvements under extreme scarcity and support stronger performance as supervision grows.
%R 10.63317/2545w4h6dty8
%U https://aclanthology.org/2026.slide-1.20/
%U https://doi.org/10.63317/2545w4h6dty8
%P 224-236
Markdown (Informal)
[Semantic, Syntactic, Lexical: What Makes QA Augmentation Work in Limited Quantity?](https://aclanthology.org/2026.slide-1.20/) (Rachmat et al., SLiDE 2026)
ACL