Natural language processing in an intelligent writing strategy tutoring system

Type: Evidence | Proposition: A: Learning | Polarity: | Sector: | Country:

This study assesses the potential of natural language processing to predict human ratings of essay quality. Past studies have demonstrated that indicators such as word frequency are significant predictors of human judgments of essay quality. This study extends earlier work by using a larger data sample and an expanded set of indicators. These indicators increased accuracy and, more importantly, offer a way of providing more meaningful feedback in the context of a writing tutoring system. Students can be taught strategies to help them to generate more text (for example, strategies to facilitate free writing, planning, drafting, and elaboration), and such strategies improve their essay quality. The higher quality essays included more phrasal constructions – they were more likely to include examples and make contrasts between ideas. In their concluding paragraphs, their authors were more likely to include concluding statements (i.e., in conclusion), statements of fact (i.e., it is), and reasons (i.e., because). The Writing Pal is an intelligent tutoring system that provides writing strategy training that is aligned with these findings. It teaches students strategies for drafting and improving body and conclusion paragraphs. For instance, students are taught how to identify and edit evidence in the body of the essay that is too speculative rather than fact based and objective. Similarly, students are taught to write conclusions that summarise key arguments without presenting additional or new evidence.
The focus of this paper is on natural language processing, rather than on the gains of learners. However, it implies that using the findings of the study to inform an intelligent tutoring system has improved essay quality.

Citation: McNamara, Danielle S, Crossley, Scott A, & Roscoe, Rod. (2013). Natural language processing in an intelligent writing strategy tutoring system. Behavioral Research, 45, 499-515. | Url:

Leave a Reply