A Comparative Study of the Feedback of Different Large Language Models on the Lexical Richness in English Writing Written by Non-English Major College Students
Based on the four dimensions of lexical richness proposed by Read(2000),this article compares the recom-mended modifications and revised compositions output by the Doubao Model,GPT-4o,and Ernie Bot on lexical richness,and analyzes the causes of lexical errors.By using TAALED,RANGE32,and AntConc4.2.4 to compare the differences in lexical variation,lexical complexity,and lexical density between the revised compositions,students can make targeted modifications and ultimately improve their English writing quality.The results are as follows:As regards lexical variation and lexical com-plexity,the numerical value of the compositions revised by Ernie Bot is higher than the Doubao Model and GPT-4o;There is no significant difference in terms of lexical density between Ernie Bot and GPT-4o,and the numerical value of both are higher than the Doubao Model.The recommended modifications concerning lexical errors made by the three large language models all tend to focus on grammatical errors and spelling errors,with the Doubao Model also pointing out errors in collocation.
Writing teachingFeedbackLarge language modelLexical richness
万方数据
维普
