On the Use of BERT for Automated Essay Scoring: Joint Learning of Multi-Scale Essay Representation

TL;DR

This paper introduces a multi-scale BERT-based model for Automated Essay Scoring that leverages joint learning, multiple losses, and transfer learning, achieving near state-of-the-art results and demonstrating strong generalization to other datasets.

Contribution

It proposes a novel multi-scale essay representation for BERT, jointly learned with multiple losses and transfer learning, advancing automated essay scoring methods.

Findings

Achieves near state-of-the-art results on the ASAP dataset.

Generalizes well to the CommonLit Readability Prize dataset.

Demonstrates the effectiveness of multi-scale representation and joint learning.

Abstract

In recent years, pre-trained models have become dominant in most natural language processing (NLP) tasks. However, in the area of Automated Essay Scoring (AES), pre-trained models such as BERT have not been properly used to outperform other deep learning models such as LSTM. In this paper, we introduce a novel multi-scale essay representation for BERT that can be jointly learned. We also employ multiple losses and transfer learning from out-of-domain essays to further improve the performance. Experiment results show that our approach derives much benefit from joint learning of multi-scale essay representation and obtains almost the state-of-the-art result among all deep learning models in the ASAP task. Our multi-scale essay representation also generalizes well to CommonLit Readability Prize data set, which suggests that the novel text representation proposed in this paper may be a new and effective choice for long-text tasks.