Toward Using Text Summarization for Essay-Based Feedback
Jill C. Burstein, Ets Technologies, Daniel Marcu · 2000
We empirically study the impact of using automatically generated summaries in the context of electronic essay rating. Our results indicate that 40% and 60% discourse-based essay summaries improve the performance of the topical analysis module of e-rater. E-rater is a system that electronically scores GMAT essays. We envision using automatically generated essay summaries for instructional feedback, as a supplement to the e-rater score. 1. Introduction Educational Testing Service (ETS) has been successfully using their automated essay scoring system, e-rater, as one of the two readers for the Graduate Management Admissions Test (GMAT) Analytical Writing Assessment (AWA). The system was designed to score essays based on holistic scoring guides (scoring rubrics) - see http://www.gmat.org for sample scoring rubrics. Holistic scoring guides instruct the human reader to assign an essay score based on the general quality of writing characteristics in an essay. For instance, the reader is to ...