Current cross-prompt automatic essay scoring (AES) systems are primarily concerned with obtaining shared knowledge specific to the target prompt by using the source and target prompt essays. However, it may not be feasible in practical situations because the target prompt essays may not be available as training data. When constructing a model solely from source prompt essays, its capacity to generalize to the target prompt may be hindered by significant discrepancies among different prompt essays. In this study, a novel learning framework for cross-prompt AES is proposed in order to capture more general knowledge across prompts and improve the model’s capacity to distinguish between writing levels. To acquire generic knowledge across different prompts, a primary model is trained via meta learning with all source prompt essays. To improve the model’s ability to differentiate writing levels, we present a level-aware learning strategy consisting of a general scorer and three level scorers for low-, middle-, and high-level essays. Then, we introduce a contrastive learning strategy to bring the essay representation of the general scorer closer to its corresponding level representation and far away from the other two levels, thereby improving the system’s ability to differentiate writing levels as well as boosting scoring performance. Experimental results on public datasets illustrate the efficacy of our method.

Current cross-prompt automated essay scoring (AES) is a challenging task due to the large discrepancies between different prompts, such as different genres and expressions. The main goal of current cross-prompt AES systems is to learn enough shared features between the source and target prompts to grade well on the target prompt. However, because the features are captured based on the original prompt representation, they may be limited by being extracted directly between essays. In fact, when the representations of two prompts are more similar, we can gain more shared features between them. Based on this motivation, in this paper, we propose a learning strategy called “prompt-mapping” to learn about more consistent representations of source and target prompts. In this way, we can obtain more shared features between the two prompts and use them to better represent the essays for the target prompt. Experimental results on the ASAP++ dataset demonstrate the effectiveness of our method. We also design experiments in different settings to show that our method can be applied in different scenarios. Our code is available at https://github.com/gdufsnlp/PMAES.

Deep learning models for automatic readability assessment generally discard linguistic features traditionally used in machine learning models for the task. We propose to incorporate linguistic features into neural network models by learning syntactic dense embeddings based on linguistic features. To cope with the relationships between the features, we form a correlation graph among features and use it to learn their embeddings so that similar features will be represented by similar embeddings. Experiments with six data sets of two proficiency levels demonstrate that our proposed methodology can complement BERT-only model to achieve significantly better performances for automatic readability assessment.