Using synthetic sentences for developing a corpus-based feedback service for student writing

Runte, Maren; Mahlow, Cerstin; Ulasik, Malgorzata Anna; Cho, Sooyeon

Publication type:	Conference paper
Type of review:	Peer review (abstract)
Title:	Using synthetic sentences for developing a corpus-based feedback service for student writing
Authors:	Runte, Maren Mahlow, Cerstin Ulasik, Malgorzata Anna Cho, Sooyeon
et. al:	No
Conference details:	SIG Writing, Paris Nanterre University, France, 26-28 June 2024
Issue Date:	26-Jun-2024
Language:	English
Subjects:	Corpus linguistics; Student writing; Annotation; Move; Step
Subject (DDC):	410.285: Computational linguistics 808: Rhetoric and writing
Abstract:	To support students when writing their bachelor thesis, early and immediate feedback on structure and linguistic realization of required elements has been proven useful. Feedback should thus influence the writing process and help students produce appropriate products. Automatic feedback systems usually rely on models derived from scientific articles written by established researchers (e.g., Weder 2015) which also serve as examples shown to students. However, student writing is not on the same level as expert writing as students are still in the process to learn and master relevant competences. Automated services, intended to support students learning to write scientific introductions, should thus be based on authentic student writing to offer appropriate feedback. These services should offer feedback on drafts, let writers revise their texts and give feedback with respect to improvements, thus guiding the writing process.Within the project «Digital Literacy in University Contexts», we develop such a service, based on over 5000 student theses written in German from several study fields. So far, we manually annotated 1034 introductions using an adapted scheme (Runte et al. 2022) based on Weder (2015). The extracted 9032 step-annotated sentences served both as a resource for corpus-linguistic investigations and training material for machine learning algorithms. The corpus-linguistic findings were used as additional features both for training models and to define feedback categories to be shown to students. As a result of our corpus-linguistic exploration, we could classify two kinds of keywords: topic-specific and procedural keywords and keyword phrases. Additionally, we identified specific linguistic features associated with certain steps. These findings were used twofold: (1) to train recommender systems for semi-automatic annotation allowing us to annotate a larger amount of introductions needed to train the automatic step detection system, and (2) as input for the creation of 1268 synthetic sentences with the API of OpenAI’s GPT-3 to arrive at a decent amount of balanced training data for our models. Additionally, we use the phrases and features to provide students with real-world examples from fellow students.
URI:	https://digitalcollection.zhaw.ch/handle/11475/31099
Fulltext version:	Published version
License (according to publishing contract):	Not specified
Departement:	Applied Linguistics
Organisational Unit:	Institute of Language Competence (ILC)
Published as part of the ZHAW project:	Digital Literacy im Hochschulkontext (DigLit)
Appears in collections:	Publikationen Angewandte Linguistik

Files in This Item:

There are no files associated with this item.

Show full item record

Runte, M., Mahlow, C., Ulasik, M. A., & Cho, S. (2024, June 26). Using synthetic sentences for developing a corpus-based feedback service for student writing. SIG Writing, Paris Nanterre University, France, 26-28 June 2024.

Runte, M. et al. (2024) ‘Using synthetic sentences for developing a corpus-based feedback service for student writing’, in SIG Writing, Paris Nanterre University, France, 26-28 June 2024.

M. Runte, C. Mahlow, M. A. Ulasik, and S. Cho, “Using synthetic sentences for developing a corpus-based feedback service for student writing,” in SIG Writing, Paris Nanterre University, France, 26-28 June 2024, Jun. 2024.

RUNTE, Maren, Cerstin MAHLOW, Malgorzata Anna ULASIK und Sooyeon CHO, 2024. Using synthetic sentences for developing a corpus-based feedback service for student writing. In: SIG Writing, Paris Nanterre University, France, 26-28 June 2024. Conference paper. 26 Juni 2024

Runte, Maren, Cerstin Mahlow, Malgorzata Anna Ulasik, and Sooyeon Cho. 2024. “Using Synthetic Sentences for Developing a Corpus-Based Feedback Service for Student Writing.” Conference paper. In SIG Writing, Paris Nanterre University, France, 26-28 June 2024.

Runte, Maren, et al. “Using Synthetic Sentences for Developing a Corpus-Based Feedback Service for Student Writing.” SIG Writing, Paris Nanterre University, France, 26-28 June 2024, 2024.