Students’ Composition Evaluation Model Based on Natural Language Processing Algorithm

Abstract

It is subjective, time consuming and labor intensive to evaluate students’ compositions. Use of natural language processing (NLP) technology effectively improves the evaluation efficiency and reduces the burden on teachers. In order to overcome the problems of traditional models, such as over-fitting and poor generalization ability, this research studied a students’ composition evaluation model based on an NLP algorithm. A students’ composition evaluation model based on a multi-task learning framework was proposed, which completed three sub-tasks simultaneously using the NLP algorithm. Three different encoding methods were used; namely, convolutional neural network (CNN), recurrent neural network (RNN), and long short-term memory (LSTM), which captured text information from multiple perspectives. A new pairing pre-training mode was built, which aimed to help build an NLP-based students’ composition evaluation model based on the multi-task learning framework, thus alleviating the deviation caused by excessive correlation. The experimental results verified that the constructed model and the proposed method were effective.