Automated Essay Scoring Using Transformer Models

Automated essay scoring (AES) is gaining increasing attention in the education sector as it significantly reduces burden of manual and allows ad hoc feedback for learners. Natural language processing based on machine learning has been shown to be particularly suitable text classification AES. While many machine-learning approaches AES still rely a bag words (BOW) approach, we consider transformer-based approach this paper, compare its performance logistic regression model BOW discuss their differences. The analysis 2088 email responses problem-solving task that were manually labeled terms politeness. Both transformer models considered outperformed without any hyperparameter tuning regression-based model. We argue that, tasks such politeness classification, significant advantages, while suffers from not taking word order into account reducing stem. Further, show how can help increase accuracy human raters, provide detailed instruction implement one’s own purposes.