1. ZHAW digitalcollection
2. Angewandte Linguistik
3. Publikationen
4. Publikationen Angewandte Linguistik

Publication type:	Conference paper
Type of review:	Peer review (abstract)
Title:	Automated sentence-centric modeling of writing
Authors:	Ulasik, Malgorzata Anna Mahlow, Cerstin
et. al:	No
Conference details:	SIG Writing, Paris Nanterre University, France, 26-28 June 2024
Issue Date:	28-Jun-2024
Language:	English
Subjects:	Keystroke-logging; Writing process; Sentence; Linguistic modeling; Writing model
Subject (DDC):	808: Rhetoric and writing
Abstract:	Sentences are typically constructed “from proposed sentence parts in a complex activity involving idea generation, evaluation, planning, and reading the text produced so far” (Hayes 2009, p. 2, in reference to the study in Kaufer et al. 1986). In this paper, we propose new methods allowing for tracking this process (the part visible on the surface) by using keystroke logging data. These methods are based on sentence-centric modeling of writing. We propose two models: (1) The first model is based on the notion of text history (Mahlow et al. 2022). Modeling text production and revision is understood as sentence-driven process. We project text transformations leading from one text version to another in the text history on sentences produced so far and pause bursts. When doing so, we distinguish between two transformation stages: initial and revision draft of the text (see Baaijen et al. 2012). This model reflects how the sentence production process is interrupted by writing mode switches, and pauses, and impacted by non-linearity. (2) The foundation for the second model is the concept of sentence history (Mahlow et al. 2022). The model represents how sentences evolve during text production. This model reflects the sentence production cycle, from transcription of an initial idea (or parts of it) through a complete sentence to a revised sentence version (potentially over several revisions) as then present in the final product. We present a working implementation of both models that allow for both testing the validity of our theoretical ideas and analyzing writing process data fully automatically towards linguistic explainability. We use a small dataset of keystroke logging data to illustrate our models. With this example case study, we show which potential insights a sentence-centric analysis can provide and point out challenges related to sentence-centric modeling in general and when performed automatically by processing keystroke-logs.
URI:	https://digitalcollection.zhaw.ch/handle/11475/31098
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:	SPPC: Swiss Process–Product Corpus of Student Writing Development
Appears in collections:	Publikationen Angewandte Linguistik

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.