Performance Comparison of Automated Essay Graders Based on Various Language Models
Xinfeng Ye, Sathiamoorthy Manoharan · 2021
Manual grading of essays is a time-consuming task. Students in large classes may therefore get delayed feedback and/or subpar feedback on their essay submissions. This has a negative impact on the students' learning experience. Many automated graders have been developed to relieve instructors from time-consuming grading tasks and providing timely feedback to students. In recent years, research in deep learning has resulted in several language models. This paper compares the performance of several automated graders that are built on top of three language models: BERT, RoBERTa and DeBERTa.