Automated essay scoring using Bayes' theorem

Abstract

Two Bayesian models for text classification from the information science field were extended and applied to student produced essays. Both models were calibrated using 462 essays with two score points. The calibrated systems were applied to 80 new, pre-scored essays with 40 essays in each score group. Manipulated variables included the two models; the use of words, phrases and arguments; two approaches to trimming; stemming; and the use of stopwords. While the text classification literature suggests the need to calibrate on thousands of cases per score group, accuracy of over 80% was achieved with the sparse dataset used in this study.