Narrative writing from users-in-the-wild

Abstract

In this chapter, we discuss the use of DocuScope as a computational tool for measuring ‘distances’ between corpora representing different contexts of writing with partially overlapping goals. In prior work (Beigman Klebanov et al., 2019) we found that there is a quantifiable sense in which argumentative essays written as part of higher ed coursework, argument essays written for a high-stakes examination, and New York Times OpEds belong to the same genre, as realized through a distribution of DocuScope categories, even though the specifics of each context are different. We extend this work in two ways: (a) we focus on narrative writing, and (b) we explore DocuScope as a basis for feedback to the writer about the extent to which a draft shows a distribution of rhetorical categories that is “close” to what one would expect for the target genre. In particular, we consider essays submitted through an online writing assistant tool Writing Mentor that are marked as narratives by their authors – a designation that we attempt to verify using DocuScope.