SEDNN: Shared and enhanced deep neural network model for cross-prompt automated essay scoring

Abstract Most existing studies on Automated Essay Scoring (AES) focused on a single prompt, and only a few studies have addressed the problem of cross-prompt AES. This paper addresses the key question of how to extract more transferable rating knowledge from multiple source prompts. Different from the common approach which extracts an invariant part among all prompts, we propose to fuse all the data from the source prompts and to extract their shared knowledge with the target prompt, including prompt-independent features and some prompt-dependent features. We show that more rating knowledge can be extracted and transferred to the target prompt using this strategy. The transferred model is then used to generate a set of pseudo training data on the target prompt, with which another enhanced model incorporating more prompt-dependent features is built for the target prompt. Experiments show that this enhanced model can further improve the final performance and it outperforms the state-of-the-art methods on cross-prompt AES.