Evaluation Method of Writing Fluency Based on Machine Learning Method

Abstract

In light of society's rapid advancement, more and more people worldwide are placing importance on education. There are several domains in China where the importance of writing exceeds the importance of reading, listening, or speaking. It has been shown that many Chinese students commit grammar problems if they are writing an article. Several researchers attempted to determine students' writing talents in terms of amount and complexity, on the one side, and then also focused on identifying conclusions on the accuracy, the organization of ideas, and the barriers to fluent writing via qualitative data gathering approaches. This research uses a machine learning technique to measure students' writing fluency. Writing fluency capabilities can be predicted using a novel adaptive generative adversarial network-based deep support vector machine (AGAN-DSVM) technique. The trace-oriented approach can be used to examine the features like accuracy, syntactic complexity, and organization of ideas aspects. The prediction rate of lexical complexity and sentence complexity of our proposed method achieves 90 and 95%, respectively. Plots created with origin's graphing tool display the results of a comparison between the proposed approach and several other ways already in use. The proposed method is evaluated and compared using several different metrics, including the accuracy dimension, syntactic complexity dimension, organization of ideas dimension, distributions of the mistakes in the text, lexical complexity, sentence complexity, essay particularities, and comparison of accuracy, F1 score, and syntactic complexity.