Chapter 3: Integrating Natural Language Processing for Writing Assessment: Writing Trait Model

Writing varies across multiple dimensions. Some of these dimensions affect writing quality (Diederich et al., 1961) and must therefore be addressed during instruction. Teachers often present sample texts that model such traits (Gallagher, 2011) and train students to use them to evaluate and revise their writing (Culham, 2003). However, analyzing a text in terms of multiple traits can be both laborious and time-consuming (Weigle, 2002). This fact suggests that there may be significant instructional advantages to producing trait scores using automated writing evaluation (AWE) systems, since AWE feedback can reduce the burden of scoring for teachers and enable fast feedback loops that supply students with personalized feedback.