Dataset adaption and evaluation on NLP models for automated grading system

Abstract

With the development of artificial intelligence in recent years, NLP (Natural Language Processing) has become increasingly mature, and one of its applica tions includes AGS (Automated Grading System). AGS plays a very important role in today’s educational learning by assigning a specific score to a given response to a specific question. And how to improve the accuracy of AGS be comes a crucial issue. One of them is how to process at the data level to make the results better, and the other is how to choose a better language model to achieve better results. In this dissertation, we compare the data and network structure to illustrate how to train a better AGS system.