Intelligent Scoring of English Composition by Machine Learning from the Perspective of Natural Language Processing

To eliminate the shortcomings of the current English composition scoring, this study intends to propose an automatic scoring model by artificial intelligence (AI) technology—machine learning (ML)—from the perspective of natural language processing (NLP). According to the sentence judgment criteria and scoring criteria of English composition, and with a combination of the n-gram model and decision tree (DT) algorithm as technical support, a composition scoring system is established with the content of the composition as the extraction feature. The random forest algorithm is taken as the training algorithm of the model. By using different maximum feature extraction numbers to test the model using the random forest training method, it is found that when the number of features is 11, the model can play the best prediction performance. Similarly, different learning rate values are used to verify the composition scoring rate of the model. It is found that when the learning rate is 6E-6, the model can play the best scoring performance. Therefore, it is believed that this automatic composition scoring model can be applied to English composition scoring in schools, and some theoretical concepts are put forward for intelligent education. The proposed composition scoring system by N-gram model and DT algorithm and trained by the random forest algorithm has high performance, which can be applied in the automatic scoring of students’ English composition in schools, and provides a basis for the educational applications of ML under AI.