@inproceedings{zhang-2026-silent,
title = "Silent Structured-Content Corruption in Document Parsing: A Span-Typed Diagnostic and a Released Cross-Parser Resource",
author = "Zhang, Ruiyang",
editor = "T.Y.S.S., Santosh and
Agarwal, Amit and
Gupta, Vivek and
Kementchedjhieva, Yova and
Garimella, Aparna and
Hou, Yufang",
booktitle = "The Proceedings of Workshop on Document Intelligence and Understanding",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.docinsights-1.19/",
pages = "273--281",
ISBN = "979-8-89176-477-4",
abstract = "Document-AI pipelines increasingly begin by converting a PDF to text with a parser{---}classical extractors, OCR, layout models, or vision-language models{---}before any retrieval or reasoning. Overall-quality metrics (edit distance, TEDS) score these parsers as if all characters mattered equally. They do not: a corrupted number, date, identifier, or table cell changes meaning while barely moving a similarity score. We isolate this failure with a span-typed diagnostic over six parsers, ten structured span types, and two conditions (digital and scanned), on a controlled corpus with exact ground truth plus a real arXiv subset. Three findings. (i) Preservation is highly parser- and type-dependent: the best parser preserves 99.4{\%} of spans while others fail specific types catastrophically (Docling drops markup to 6.0{\%}; MarkItDown returns nothing on scanned pages), so ``which parser'' is a per-span-type risk decision, not a single score. (ii) Document-level text similarity does not predict span loss (Spearman {\ensuremath{\rho}} {\ensuremath{\leq}} 0.46 across all parsers and conditions; {\ensuremath{\approx}} 0 for VLMs), so standard metrics are blind to it. (iii) Downstream, corrupted spans are dangerous precisely when the parse looks successful: among questions whose evidence span was corrupted but whose document still returned text, 96.5{\%} are answered with a confident wrong value rather than an abstention{---}an undetectable error{---}whereas total parse failure is comparatively safe because it forces an abstention. We release the resource behind these numbers: 566,904 per-span verdicts over 660 documents and 7,560 parser outputs, under CC-BY, with an independent verifier (49,720 checks) that re-derives every reported number."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhang-2026-silent">
<titleInfo>
<title>Silent Structured-Content Corruption in Document Parsing: A Span-Typed Diagnostic and a Released Cross-Parser Resource</title>
</titleInfo>
<name type="personal">
<namePart type="given">Ruiyang</namePart>
<namePart type="family">Zhang</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>The Proceedings of Workshop on Document Intelligence and Understanding</title>
</titleInfo>
<name type="personal">
<namePart type="given">Santosh</namePart>
<namePart type="family">T.Y.S.S.</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Amit</namePart>
<namePart type="family">Agarwal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vivek</namePart>
<namePart type="family">Gupta</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yova</namePart>
<namePart type="family">Kementchedjhieva</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Aparna</namePart>
<namePart type="family">Garimella</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yufang</namePart>
<namePart type="family">Hou</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">979-8-89176-477-4</identifier>
</relatedItem>
<abstract>Document-AI pipelines increasingly begin by converting a PDF to text with a parser—classical extractors, OCR, layout models, or vision-language models—before any retrieval or reasoning. Overall-quality metrics (edit distance, TEDS) score these parsers as if all characters mattered equally. They do not: a corrupted number, date, identifier, or table cell changes meaning while barely moving a similarity score. We isolate this failure with a span-typed diagnostic over six parsers, ten structured span types, and two conditions (digital and scanned), on a controlled corpus with exact ground truth plus a real arXiv subset. Three findings. (i) Preservation is highly parser- and type-dependent: the best parser preserves 99.4% of spans while others fail specific types catastrophically (Docling drops markup to 6.0%; MarkItDown returns nothing on scanned pages), so “which parser” is a per-span-type risk decision, not a single score. (ii) Document-level text similarity does not predict span loss (Spearman \ensuremathρ \ensuremathłeq 0.46 across all parsers and conditions; \ensuremath\approx 0 for VLMs), so standard metrics are blind to it. (iii) Downstream, corrupted spans are dangerous precisely when the parse looks successful: among questions whose evidence span was corrupted but whose document still returned text, 96.5% are answered with a confident wrong value rather than an abstention—an undetectable error—whereas total parse failure is comparatively safe because it forces an abstention. We release the resource behind these numbers: 566,904 per-span verdicts over 660 documents and 7,560 parser outputs, under CC-BY, with an independent verifier (49,720 checks) that re-derives every reported number.</abstract>
<identifier type="citekey">zhang-2026-silent</identifier>
<location>
<url>https://aclanthology.org/2026.docinsights-1.19/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>273</start>
<end>281</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Silent Structured-Content Corruption in Document Parsing: A Span-Typed Diagnostic and a Released Cross-Parser Resource
%A Zhang, Ruiyang
%Y T.Y.S.S., Santosh
%Y Agarwal, Amit
%Y Gupta, Vivek
%Y Kementchedjhieva, Yova
%Y Garimella, Aparna
%Y Hou, Yufang
%S The Proceedings of Workshop on Document Intelligence and Understanding
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-477-4
%F zhang-2026-silent
%X Document-AI pipelines increasingly begin by converting a PDF to text with a parser—classical extractors, OCR, layout models, or vision-language models—before any retrieval or reasoning. Overall-quality metrics (edit distance, TEDS) score these parsers as if all characters mattered equally. They do not: a corrupted number, date, identifier, or table cell changes meaning while barely moving a similarity score. We isolate this failure with a span-typed diagnostic over six parsers, ten structured span types, and two conditions (digital and scanned), on a controlled corpus with exact ground truth plus a real arXiv subset. Three findings. (i) Preservation is highly parser- and type-dependent: the best parser preserves 99.4% of spans while others fail specific types catastrophically (Docling drops markup to 6.0%; MarkItDown returns nothing on scanned pages), so “which parser” is a per-span-type risk decision, not a single score. (ii) Document-level text similarity does not predict span loss (Spearman \ensuremathρ \ensuremathłeq 0.46 across all parsers and conditions; \ensuremath\approx 0 for VLMs), so standard metrics are blind to it. (iii) Downstream, corrupted spans are dangerous precisely when the parse looks successful: among questions whose evidence span was corrupted but whose document still returned text, 96.5% are answered with a confident wrong value rather than an abstention—an undetectable error—whereas total parse failure is comparatively safe because it forces an abstention. We release the resource behind these numbers: 566,904 per-span verdicts over 660 documents and 7,560 parser outputs, under CC-BY, with an independent verifier (49,720 checks) that re-derives every reported number.
%U https://aclanthology.org/2026.docinsights-1.19/
%P 273-281
Markdown (Informal)
[Silent Structured-Content Corruption in Document Parsing: A Span-Typed Diagnostic and a Released Cross-Parser Resource](https://aclanthology.org/2026.docinsights-1.19/) (Zhang, DocInsights 2026)
ACL