基于文本复杂度的ChatGPT和研究生写作的语言差异研究

The Comparative Analysis of Compositions Written by Postgraduates and ChatGPT and Its Implications on Second Language Writing

摘要: 作为巨型语言模型，ChatGPT的横空出世给教育领域带来了深远的影响。本研究旨在探讨ChatGPT与二语学习者写作文本的差异。通过运用Coh-Metrix文本分析工具与SPSS数据分析软件，对比了ChatGPT与20位应用语言学专业研究生的作文文本。结果表明，在词法复杂度方面，学生习作与ChatGPT在词汇量上无显著差异，但学生词汇密度更高；在句法复杂度上，学生倾向于使用短句，句法使用相似度高于ChatGPT。ChatGPT能有效避免常见写作错误，但缺乏主观情感与修辞效果。因此，ChatGPT虽强大，但不可完全依赖，其高效利用取决于使用者的指令。本研究为AI在二语写作中的应用提供了初步参考。

Abstract: The introduction of ChatGPT, a colossal language model, has significantly reshaped educational landscapes, This investigation sets out to elucidate disparities between texts generated by ChatGPT and those authored by second language learners. Leveraging the Coh-Metrix analytical tool and SPSS for rigorous data examination, a comparative analysis was conducted between compositions produced by ChatGPT and 20 applied linguistics graduates. Findings revealed parity in lexical breadth between student work and ChatGPT, albeit with students exhibiting heightened vocabulary density. Syntactically, students favored concise structures, presenting a higher syntactic similarity index compared to ChatGPT’s more diverse usage. ChatGPT excelled in averting typical writing mistakes, yet it lacked the emotive depth and rhetorical flourish characteristic of human composition, Consequently while potent, ChatGPT’s utility is not absolute and hinges on adept user guidance, thereby offering a foundational premise for Al integration in second language writing pedagogy.