Research on Automatic Essay Scoring of Composition Based on CNN and OR

Manual scoring may be affected by unconscious factors such as subjective judgment of the reviewer, which may lead to the deviation of scores. Establishing an objective, systematic automatic comment system is of great significance to solve this problem. Most of the existing automatic scoring systems are based on a single neural network or based on feature extraction. The accuracy of the formers need to be improved, and the latter require a lot of time to manually extract features. After exploring and analyzing convolutional neural networks (CNN) and ordinal regression (OR), we propose an automatic essay scoring approach based on a combination of CNN and OR for the characteristics of automatic scoring mechanism. Through the deep learning framework Keras to realize the designing, the experimental results demonstrate that the proposed model has a great improvement on the accuracy and efficiency of the automatic essay scoring than existing methods.