Use of Data Augmentation in Automated Essay Scoring Training Data

• In Event: Automated Scoring II

Abstract

Automated Essay Scoring (AES) requires costly human scored responses to train an engine. Our study explores how data augmentation (DA) method and insertion rate impacts AES model accuracy. Our results suggest that DA methods may improve model performance with as few as 10 generated examples.

Authors

• Justin O Barber, Pearson

• Edward W Wolfe, Pearson