On Curating HTR Training Datasets for Romanian Language with use of Transcribathon Tool

This paper presents a workflow for HTR dataset generation in Romanian using Transcribathon’s Correct HTR feature. Leveraging citizen-science transcriptions aligned with Transkribus outputs, our case study on Jurnalul lui Dumitru Nistor reduced CER from 15.26% to 0.13%. The approach enables efficient dataset curation and supports scalable model development in low-resource languages.

Authors

Sergiu Gordea (ORCID: https://orcid.org/0000-0003-4385-7533)
Frank Drauschke (ORCID: https://orcid.org/0009-0009-1432-8836)
George Christian Cotea
Tom Vanallemeersch
Jasmin Salesevic
Marthe Lamote

Institutions

AIT Austrian Institute of Technology GmbH (AT)
CrossLang (Belgium) (BE)
Digital Science (United States) (US)

Publication Details

Journal: Zenodo (CERN European Organization for Nuclear Research)
Published: 2026-02-18
DOI: https://doi.org/10.5281/zenodo.18890949
Primary Topic: Mathematics, Computing, and Information Processing
Type: article
Field-Weighted Citation Impact: 0.00

Controls

|||

ALL TIME

JAN

FEB

MAR

APR

MAY

JUN

JUL

AUG

SEP

OCT

article

On Curating HTR Training Datasets for Romanian Language with use of Transcribathon Tool

Sergiu Gordea, Frank Drauschke, George Christian Cotea, Tom Vanallemeersch et al.

Feb 18, 2026

Zenodo (CERN European Organization for Nuclear Research)

Mathematics, Computing, and Information Processing

article

On Curating HTR Training Datasets for Romanian Language with use of Transcribathon Tool

Sergiu Gordea, Frank Drauschke, George Christian Cotea, Tom Vanallemeersch, Jasmin Salesevic, Marthe Lamote

February 18, 2026article en

Abstract

This paper presents a workflow for HTR dataset generation in Romanian using Transcribathon’s Correct HTR feature. Leveraging citizen-science transcriptions aligned with Transkribus outputs, our case study on Jurnalul lui Dumitru Nistor reduced CER from 15.26% to 0.13%. The approach enables efficient dataset curation and supports scalable model development in low-resource languages.

Zenodo (CERN European Organization for Nuclear Research)

AIT Austrian Institute of Technology GmbH (AT), CrossLang (Belgium) (BE), Digital Science (United States) (US)

Openalex Percentile: Top 31%

Mathematics, Computing, and Information Processing

DOI Open Access PDF OpenAlex

AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

On Curating HTR Training Datasets for Romanian Language with use of Transcribathon Tool — Sergiu Gordea, Frank Drauschke, et al. · Zenodo (CERN European Organization for Nuclear Research) (2026) | TGRS Research Map | TGRS