@inproceedings{sindelar-etal-2026-training,
title = "Training data generation for context-dependent rubric-based short answer grading",
author = "{\v{S}}indel{\'a}{\v{r}}, Pavel and
Pr{\'a}{\v{s}}il, Filip and
Slivka, D{\'a}vid and
Bouma, Christopher and
Bojar, Ondrej",
editor = {Barth, Florian and
Du, Keli and
Calvo Tello, Jos{\'e} and
Gen{\^e}t, Philippe and
Lendvai, Piroska and
Sch{\"o}ch, Christof and
Trippel, Thorsten},
booktitle = "Proceedings of Leveraging Derived Text Formats to Unlock Copyrighted Collections for Open Science ({DTF}) @ {LREC} 2026",
month = jun,
year = "2026",
address = "Palma, Mallorca (Spain)",
publisher = "ELRA Language Resources Association (ELRA)",
url = "https://aclanthology.org/2026.dtf-1.6/",
doi = "10.63317/3tt9vin8tdc4",
pages = "44--50",
abstract = "Every four years, the PISA test is administered by the OECD to test the knowledge of teenage students worldwide and allow for comparisons of educational systems. However, having to avoid language differences and annotator bias makes the grading of student answers challenging. For these reasons, it would be interesting to consider methods of automatic student answer grading. To train some of these methods, which require machine learning, or to compute parameters or select hyperparameters for those that do not, a large amount of domain-specific data is needed. In this work, we explore a small number of methods for creating a large-scale training dataset using only a relatively small confidential dataset as a reference, leveraging a set of very simple derived text formats to preserve confidentiality. Using the proposed methods, we successfully created three surrogate datasets that are, at the very least, superficially more similar to the reference dataset than a straightforward result of prompt-based generation. Early experiments suggest one of these approaches might also lead to improved training of automatic answer grading models."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="sindelar-etal-2026-training">
<titleInfo>
<title>Training data generation for context-dependent rubric-based short answer grading</title>
</titleInfo>
<name type="personal">
<namePart type="given">Pavel</namePart>
<namePart type="family">Šindelář</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Filip</namePart>
<namePart type="family">Prášil</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dávid</namePart>
<namePart type="family">Slivka</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Christopher</namePart>
<namePart type="family">Bouma</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ondrej</namePart>
<namePart type="family">Bojar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-06</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of Leveraging Derived Text Formats to Unlock Copyrighted Collections for Open Science (DTF) @ LREC 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Florian</namePart>
<namePart type="family">Barth</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Keli</namePart>
<namePart type="family">Du</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">José</namePart>
<namePart type="family">Calvo Tello</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Philippe</namePart>
<namePart type="family">Genêt</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Piroska</namePart>
<namePart type="family">Lendvai</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Christof</namePart>
<namePart type="family">Schöch</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Thorsten</namePart>
<namePart type="family">Trippel</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>ELRA Language Resources Association (ELRA)</publisher>
<place>
<placeTerm type="text">Palma, Mallorca (Spain)</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Every four years, the PISA test is administered by the OECD to test the knowledge of teenage students worldwide and allow for comparisons of educational systems. However, having to avoid language differences and annotator bias makes the grading of student answers challenging. For these reasons, it would be interesting to consider methods of automatic student answer grading. To train some of these methods, which require machine learning, or to compute parameters or select hyperparameters for those that do not, a large amount of domain-specific data is needed. In this work, we explore a small number of methods for creating a large-scale training dataset using only a relatively small confidential dataset as a reference, leveraging a set of very simple derived text formats to preserve confidentiality. Using the proposed methods, we successfully created three surrogate datasets that are, at the very least, superficially more similar to the reference dataset than a straightforward result of prompt-based generation. Early experiments suggest one of these approaches might also lead to improved training of automatic answer grading models.</abstract>
<identifier type="citekey">sindelar-etal-2026-training</identifier>
<identifier type="doi">10.63317/3tt9vin8tdc4</identifier>
<location>
<url>https://aclanthology.org/2026.dtf-1.6/</url>
</location>
<part>
<date>2026-06</date>
<extent unit="page">
<start>44</start>
<end>50</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Training data generation for context-dependent rubric-based short answer grading
%A Šindelář, Pavel
%A Prášil, Filip
%A Slivka, Dávid
%A Bouma, Christopher
%A Bojar, Ondrej
%Y Barth, Florian
%Y Du, Keli
%Y Calvo Tello, José
%Y Genêt, Philippe
%Y Lendvai, Piroska
%Y Schöch, Christof
%Y Trippel, Thorsten
%S Proceedings of Leveraging Derived Text Formats to Unlock Copyrighted Collections for Open Science (DTF) @ LREC 2026
%D 2026
%8 June
%I ELRA Language Resources Association (ELRA)
%C Palma, Mallorca (Spain)
%F sindelar-etal-2026-training
%X Every four years, the PISA test is administered by the OECD to test the knowledge of teenage students worldwide and allow for comparisons of educational systems. However, having to avoid language differences and annotator bias makes the grading of student answers challenging. For these reasons, it would be interesting to consider methods of automatic student answer grading. To train some of these methods, which require machine learning, or to compute parameters or select hyperparameters for those that do not, a large amount of domain-specific data is needed. In this work, we explore a small number of methods for creating a large-scale training dataset using only a relatively small confidential dataset as a reference, leveraging a set of very simple derived text formats to preserve confidentiality. Using the proposed methods, we successfully created three surrogate datasets that are, at the very least, superficially more similar to the reference dataset than a straightforward result of prompt-based generation. Early experiments suggest one of these approaches might also lead to improved training of automatic answer grading models.
%R 10.63317/3tt9vin8tdc4
%U https://aclanthology.org/2026.dtf-1.6/
%U https://doi.org/10.63317/3tt9vin8tdc4
%P 44-50
Markdown (Informal)
[Training data generation for context-dependent rubric-based short answer grading](https://aclanthology.org/2026.dtf-1.6/) (Šindelář et al., DTF 2026)
ACL