Automatic essay scoring with recurrent neural network

As deep learning has developed rapidly in recent years, the automatic essay scoring system, based on deep learning models, has become more reliable than previous feature-based systems. Recent researchers have developed an approach based on recurrent neural networks to learn the relationship between an essay and its assigned score, without any feature engineering. In this paper, we use an ASAP essay dataset, combining feature scoring and a recurrent neural network. The results show that we can compare the result of quadratic weighted Kappa of each experience to get the best model. GloVe significantly improves the results, and feature extraction can affect the result slightly. In future work, we will apply transfer learning, one-shot learning, and adversarial inputs in our model to get better performance.