Previous
Next
Rebecca Passonneau (Penn State) – The Ideal and the Real: Why NLP is Needed for Writing Assessment
March 12, 2021
9:00 am
ZOOM Virtual Room (Link will be provided)

Rebecca Passonneau (Penn State) – The Ideal and the Real: Why NLP is Needed for Writing Assessment

The Ideal and the Real: Why NLP is Needed for Writing Assessment

For years, the National Center for Education Statistics has reported that on average, the writing skills of secondary students are below grade level, with respect to the ability to express ideas. Similar alarms are raised regarding the writing skills of college students across disciplines. Decades of research have shown that writing skills are complex, differ across disciplines, take years to achieve, and improve through practice and feedback. Teachers and instructors rely on writing to assess students’ knowledge, but lack the resources to provide more than a few writing assignments, much less to integrate writing skills into their curricula. My lab has been investigating automated techniques to support analytic writing rubrics for short essay assignments in STEM subjects, and more recently for constructed response questions.  We have also been investigating rubric reliability in the context of actual classroom use, contrasted with reliability of the same rubrics given intensive training of raters. In this talk I will give an overview of our work on automated assessment of writing, and on rubric reliability. While neither machines nor humans can assess writing sufficiently accurately and quickly in the context of classroom practice, there is great potential to partnering their complementary strengths.