Mehrebenenannotation argumentativer Lerner∗innentexte für die automatische Textauswertung

verfasst von

Sebastian Kilsbach, Sara Rezat, Nadine Michel, Rabia Karabey, Maja Stahl, Henning Wachsmuth

Abstract

This article presents a multi-level annotation approach for argumentative learner texts that was developed as part of an interdisciplinary DFG project. The project aims at the automated generation of individualized, development-promoting and learner-sensitive feedback on argumentative student texts and is situated in the field of AI-supported text production. To generate automated feedback, the first step was to manually annotate an extensive text corpus consisting of 1,320 argumentative texts written by fifth and ninth graders. This then formed the basis for the development of corresponding computational linguistic procedures. The article focuses on the special features as well as the challenges that arose in connection with the annotation of learner texts and the generation of learner-sensitive feedback. The article is structured as follows: First, the relevant computational linguistics and language didactics research findings and digital support systems for argumentative writing are outlined. In the main part, the procedure of multi-level annotation is explained in detail. Due to the methodological approach, above-average inter-annotator agreement was achieved resulting in the multi-level approach implemented being adaptable for further corpus-based studies. Finally, the results are interpreted and discussed.

Organisationseinheit(en)

Gottfried Wilhelm Leibniz Universität Hannover

Externe Organisation(en)

Universität Paderborn

Typ

Artikel

Journal

Zeitschrift fur Angewandte Linguistik

Band

2025

Seiten

102–129

ISSN

1433-9889

Publikationsdatum

13.03.2025

Publikationsstatus

Veröffentlicht

Peer-reviewed

Ja

ASJC Scopus Sachgebiete

Sprache und Linguistik, Linguistik und Sprache

Elektronische Version(en)

https://doi.org/10.1515/zfal-2025-2003 (Zugang: Offen)