Authors: H Ullrich, J Drchal, M Rýpar, H Vincourová, V Moravec
Venue: Language Resources and Evaluation 57 (4), 1571-1605
Year: 2023
Citations: 19
Links
Abstract
CsFEVER and CTKFacts introduce Czech-language resources for automated fact verification. The paper describes how claims and evidence were collected, annotated, and validated, and it highlights the challenges of balancing label quality, evidence grounding, and linguistic diversity in a lower-resource setting.
Resources
- Video: TODO
- Slides: TODO
- Code: TODO
- Dataset: TODO
Notes
This dataset work was foundational for our later shared-task systems: it gave us Czech-centric data practices, annotation lessons, and realistic expectations about retrieval and evidence quality when moving from English benchmarks to multilingual fact-checking.