Datasets:
license: cc-by-nc-sa-4.0
task_categories:
- text2text-generation
language:
- hu
tags:
- simplification
- HunSimpleNews
- news
- Hungarian
pretty_name: HunSimpleNews
size_categories:
- 1K<n<10K
HunSimpleNews
The majority of the text simplification literature in NLP focuses on simplifying sentences. From a theoretical standpoint, this approach is not entirely valid as it misses vital context required to dissolve ambiguities in meaning, references and the process itself.
HunSimpleNews is the first Hungarian text simplification corpus that includes the standard and simplified versions of whole documents. The corpus contains news article pairs taken from the Serbian Hungarian news outlet Pannon RTV. The simplification was performed manually by a journalist of the portal.
The data is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (cc-by-nc-sa-4.0) licence.
If you use the corpus in any form, please cite the following paper (in Hungarian) [bib]:
Prótár, Noémi and Nemeskey, Dávid Márk: HunSimpleNews: Az első autentikus magyar nyelvű szövegekből álló szövegegyszerűsítési korpusz.
In XXI. Magyar Számítógépes Nyelvészeti Konferencia (MSZNY2025), pages 197–218, Szeged.