Elamur Research⌕
☰
home / papers / 2026 / GENIUS Keylog Corpus - a German High School Student Corpus with Keystroke Logging Data

GENIUS Keylog Corpus - a German High School Student Corpus with Keystroke Logging Data

Written for the Research library (claude-opus-5-5) from the paper text, in our own words: the license "CC BY-NC 4.0 (ELRA, LREC proceedings)" does not allow redistributing the text itself. Original — aclanthology.org/2026.lrec-1.550.pdf

Summary

Writing research typically examines either the finished text or the way it was typed. Until now, no public resource linked keystroke-level process data to fine-grained argumentative annotation. The GENIUS Keylog corpus fills this gap with 259 argumentative essays by German students in grades 10–11. In 15 minutes, students weighed car propulsion options (battery-electric, hydrogen fuel cells, synthetic e-fuels) in light of climate change. The task replicates the DARIUS setup, but the data were collected independently.

Each key press was logged with a millisecond timestamp and character offset. This allows the full composition timeline to be rebuilt and mapped onto annotated content zones (introduction, main part, conclusion). Arguments, major claims and Toulmin elements are annotated as well. Baseline analyses compute per-zone revision ratios, graded pause categories and writing bursts. Under time pressure, these analyses show that the amount of revising, the writing pace and the distribution of time differ systematically from one essay section to another. The data, linking methodology and code are released, supporting process-aware feedback and argument mining on incomplete or in-progress texts.

Key points

  • Size and data: 273 essays were collected and 259 kept after filtering, with 233,719 keystroke events in total. Students wrote actively for about 5.2 minutes on average, out of roughly 11.2 minutes before their last keystroke.
  • Annotation: The scheme is adapted from DARIUS. Six trained annotators worked on the corpus, with at least two per essay. Agreement on content zone boundaries reached Cohen's κ = 0.83. Further layers cover arguments (topic, stance, accuracy, adequacy), major claims and Toulmin elements (claim, data, warrant, rebuttal).
  • Process metrics:
    • The revision ratio is deletions divided by productive events within a zone.
    • Pauses fall into five bins, ranging from under 500 ms to 5,000 ms or more. This extends the classic 2-second threshold (Wengelin, 2006).
    • 96% of writers worked linearly.
    • Essays averaged about 28 bursts lasting roughly 11 s each, and 79% of bursts were fluent.
  • Conclusions and grades (appendix): Only around 30 essays included a conclusion. Students who wrote one had significantly better math grades (p = .026). Longer output correlated weakly with better chemistry and physics grades (r ≈ −.16 and −.20).
  • Positioning and access: Unlike KUPA-KEYS and KLiCKe, which offer only holistic quality scores, GENIUS aligns process data with argument structure. It is available at github.com/darius-ipn/genius_keylogs.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Fields: Linguistics & languages

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.63317/4zdv5j2yj6vg

@inproceedings{schaller2026genius,
  title = {{GENIUS Keylog Corpus - a German High School Student Corpus with Keystroke Logging Data}},
  author = {Schaller, Nils-Jonathan and Jansen, Thorben and Höft, Lars and Pünjer, Hannah and Horbach, Andrea},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference},
  year = {2026},
  doi = {10.63317/4zdv5j2yj6vg},
  url = {https://doi.org/10.63317/4zdv5j2yj6vg}
}

References

17 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. R. Conijn, A. Rossetti, N. Vandermeulen, and L. Van Waes. 2025. Phase to phase: Developing an automated procedure to identify and visualize phases in writing sessions using keystroke data. Journal of Writing Research, 17(2):339–369. ↗
  2. Scott A. Crossley, Perpetual Baffour, Yu Tian, Aigner Picou, Meg Benner, and Ulrich Boser. 2022. The persuasive essays for rating, selecting, and understanding argumentative and discourse elements (persuade) corpus 1.0. Assessing Writing, 54:100667. ↗
  3. Mariëlle Leijten, Luuk Van Waes, Iris Schrijver, Sarah Bernolet, and Lieve Vangehuchten. 2019. Mapping master’s students’ use of external sources in source-based writing in l1 and l2. Studies in Second Language Acquisition, 41(3):555–582. ↗
  4. Cerstin Mahlow. 2014. Learning from errors: Systematic analysis of complex writing errors for improving writing technology. In Language Production, Cognition, and the Lexicon, pages 419–438. Springer. ↗
  5. Cerstin Mahlow. 2015. Learning from errors: Systematic analysis of complex writing errors for improving writing technology. In N. Gala, R. Rapp, and G. Bel-Enguix, editors, Language Production, Cognition, and the Lexicon, volume 48 of Text, Speech and Language Technology, pages 433–457. Springer, Cham. ↗
  6. Tanja Riemeier, Claudia Aufschnaiter, Jan Fleischhauer, and Christian Rogge. 2012. Ar-gumentationen von schülern prozessbasiert analysieren: Ansatz, vorgehen, befunde und implikationen. Zeitschrift für Didaktik der Naturwissenschaften, 18:141–180. find
  7. Nils-Jonathan Schaller, Yuning Ding, Thorben Jansen, and Andrea Horbach. 2025. Don’t score too early! evaluating argument mining models on incomplete essays. In Proceedings of the 20th Workshop on Innovative Use of NLP for Building Educational Applications (BEA 2025), pages 345–355, Vienna, Austria. Association for Computational Linguistics. find
  8. Nils-Jonathan Schaller, Andrea Horbach, Lars Ingver Höft, Yuning Ding, Jan Luca Bahr, Jennifer Meyer, and Thorben Jansen. 2024. DARIUS: A comprehensive learner corpus for argument mining in German-language essays. In Proceedings of Mahlowuistics, Language Resources and Evaluation (LREC-COLING 2024), pages 4356– 4367, Torino, Italia. ELRA and ICCL. find
  9. Ronja Schiller, Johanna Fleckenstein, Lars Höft, Andrea Horbach, and Jennifer Meyer. 2025. On the role of engagement in automated feedback effectiveness: Insights from keystroke logging. Computers & Education, 238:105386. ↗
  10. Christian Stab and Iryna Gurevych. 2014. Annotating argument components and relations in persuasive essays. In Proceedings of COLING 2014, the 25th International Conference on Computational Linguistics: Technical Papers, pages 1501–1510, Dublin, Ireland. Dublin City University and Association for Computational Linguistics. find
  11. Manfred Stede, Sara Mamprin, Andreas Peldszus, André Herzog, David Kaupat, Christian Chiarcos, and Saskia Warzecha. 2015. Handbuch Textannotation. Potsdamer Kommentarkorpus 2.0. Universitätsverlag Potsdam. find
  12. M. Talebinamvar and F. Zarrabi. 2022. Clustering students’ writing behaviors using keystroke logging: a learning analytic approach in efl writing. Language Testing in Asia, 12(6). ↗
  13. Y. Tian, S. Crossley, and L. Van Waes. 2025. The klicke corpus: Keystroke logging in compositions for knowledge evaluation. Journal of Writing Research, 17(1):23–60. ↗
  14. N. Vandermeulen, E. Van Steendam, S. De Maeyer, et al. 2024. Learning to write syntheses: the effect of process feedback and of observing models on performance and process behaviors. Reading and Writing, 37:1375–1405. ↗
  15. Georgios Velentzas, Andrew Caines, Rita Borgo, Erin Pacquetet, Clive Hamilton, Taylor Arnold, Diane Nicholls, Paula Buttery, Thomas Gaillat, Nicolas Ballier, and Helen Yannakoudakis. 2024. Logging keystrokes in writing by English learners. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), pages 10725–10746, Torino, Italia. ELRA and ICCL. find
  16. Åsa Wengelin. 2006. Examining Pauses in Writing: Theory, Methods and Empirical Data, volume 18, pages 107–130. Elsevier, United States. The information about affiliations in this record was updated in December 2015. The record was previously connected to the following departments: Linguistics and Phonetics (015010003). find
  17. M. Zhang, M. Zhu, P. Deane, and H. Guo. 2019. Identifying and comparing writing process patterns using keystroke logs. In M. Wiberg, S. Culpepper, R. Janssen, J. González, and D. Molenaar, editors, Quantitative Psychology, volume 265 of Springer Proceedings in Mathematics & Statistics, pages 459–475. Springer, Cham. find

↑ back to top