Corpus structure and sources

DE-Lite v2.1 is a subset of DE-Lite v2, which builds upon DE-Lite v1 (Jablotschkin et al. 2024). DE-Lite v2.1 comprises those files from DE-Lite v2 for which the publishers granted the rights to share the texts. We will release additional data from our in-house collection once we obtain the required licenses.

The corpus contains German language texts that have been optimised for comprehensibility for readers with low reading and/or text comprehension skills (e.g., adults with disabilities or children). It consists of two subcorpora: a monolingual and a parallel subcorpus. The monolingual subcorpus consists of texts that originally have been written in a comprehensibility-enhanced form of German (e.g., Easy German or 'Leichte Sprache'). The parallel subcorpus consists of original source texts and translations of those texts into a simpler form of German.

Some of the texts have been taken from one of the following pre-existing corpora:

Geasy (Hansen-Schirra et al. 2021)
WebCorpus (Battisti et al. 2020)
DEPlain (Stodden et al. 2023)
LeiKo (Jablotschkin & Zinsmeister 2023)

Additional texts have been collected from the web.

Annotations and metadata

Each corpus file exists in a raw text version (see subdirectory text) and in a conllu version with annotations such as parts of speech, dependencies and coreference (see subdirectory conllu). Automatic annotation was performed with UDPipe (Straka 2018) and CorPipe (Straka 2025).

All corpus files were enriched with extensive metadata (see files metadata_delite_v2_monolingual.csv and metadata_delite_v2_parallel.csv), e.g., publication date, text genre and simplification label.

See Jablotschkin (2025) for the text genre annotation guidelines.

Please note that the labels ‘easy’, ‘simple’ and ‘children’ (in German: Sprachtyp ‘Leicht’, ‘Einfach’, ‘Kinder’) are merely external classifications derived from the original publication. Different criteria were used by different publishers when determining these classifications. The LeiSA project, for example, extensively discussed whether their texts could be labelled as 'Leichte Sprache' (Easy German) in alignment with common usage practice. These discussions revealed markedly different viewpoints and, in some cases, irreconcilable arguments. Their final decision to apply the label 'Leichte Sprache' for their texts was based on considerations regarding the target audience and practical applicability, other labels for simplification could also have been chosen (Bettina Bock, p.c.).

References

Battisti, Alessia, Dominik Pfütze, Andreas Säuberli, Marek Kostrzewa, and Sarah Ebling. 2020. ‘A Corpus for Automatic Readability Assessment and Text Simplification of German’. Proceedings of The 12th Language Resources and Evaluation Conference (Marseille, France), 3302–3311. https://www.aclweb.org/anthology/2020.lrec-1.404.

Hansen-Schirra, Silvia, Jean Nitzke, and Silke Gutermuth. 2021. ‘An Intralingual Parallel Corpus of Translations into German Easy Language (Geasy Corpus): What Sentence Alignments Can Tell Us About Translation Strategies in Intralingual Translation’. In New Perspectives on Corpus Translation Studies, edited by Vincent X. Wang, Lily Lim, and Defeng Li. New Frontiers in Translation Studies. Springer Singapore, 281-298. https://doi.org/10.1007/978-981-16-4918-9_11.

Jablotschkin, Sarah. 2025. ‘Guidelines für die Annotation von Textsorten im DE-Lite-Korpus’. Version v4. Universität Hamburg. https://doi.org/10.25592/UHHFDM.17439.

Jablotschkin, Sarah, and Heike Zinsmeister. 2023. ‘LeiKo. Ein Vergleichskorpus für Leichte und Einfache Sprache’. In Neue Entwicklungen in der Korpuslandschaft der Germanistik: Beiträge zur IDS-Methodenmesse 2022, edited by Marc Kupietz and Thomas Schmidt. Korpuslinguistik und Interdisziplinäre Perspektiven auf Sprache, Bd./vol. 11. Institut für deutsche Sprache. Narr Francke Attempto, 71-88. https://doi.org/10.24053/9783823396024.

Jablotschkin, Sarah, Elke Teich, and Heike Zinsmeister. 2024. ‘DE-Lite -- a New Corpus of Easy German: Compilation, Exploration, Analysis’. In Proceedings of the Fourth Workshop on Language Technology for Equality, Diversity, Inclusion, edited by Bharathi Raja Chakravarthi, Barathi B, Paul Buitelaar, Thenmozhi Durairaj, György Kovács, Miguel Ángel, and García Cumbreras. Association for Computational Linguistics, 106-117. https://aclanthology.org/2024.ltedi-1.9/.

Stodden, Regina, Omar Momen, and Laura Kallmeyer. 2023. ‘DEplain: A German Parallel Corpus with Intralingual Translations into Plain Language for Sentence and Document Simplification’. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 16441-16463. https://doi.org/10.18653/v1/2023.acl-long.908.

Straka, Milan. 2018. ‘UDPipe 2.0 Prototype at CoNLL 2018 UD Shared Task’. Proceedings of CoNLL 2018: The SIGNLL Conference on Computational Natural Language Learning, 197–207. https://doi.org/10.18653/v1/K18-2020.

Straka, Milan. 2025. ‘CorPipe at CRAC 2025: Evaluating Multilingual Encoders for Multilingual Coreference Resolution’. Proceedings of the Eighth Workshop on Computational Models of Reference, Anaphora and Coreference, 130–139. https://doi.org/10.18653/v1/2025.crac-1.11.

This research was conducted within the scope of transfer project T1 "Information density and linguistic encoding of 'Leichte Sprache' (IDeaLite)" of SFB 1102, which has been funded by Deutsche Forschungsgemeinschaft (DFG, German Research Foundation). Project number: 232722074