流利
可靠性(半导体)
语言学
自然语言处理
心理学
人工智能
度量(数据仓库)
学术写作
计算机科学
写作评估
语言能力
认知心理学
学术英语
词汇多样性
词汇
作者
Hengzhi Hu,Kehan Du,Harwati Binti Hashim
出处
期刊:
日期:2025-06-13
卷期号:: 189-195
被引量:4
标识
DOI:10.1109/icdel65868.2025.11193569
摘要
This paper reports a practice within a Malaysian university of using ChatGPT to assess English writing proficiency through complexity, accuracy, and fluency (CAF) indices. We compared ChatGPT-generated CAF indices with expert analyses in an academic English writing test. Results showed that ChatGPT’s scores aligned well with expert ratings in syntactic complexity and fluency but exhibited inconsistencies in accuracy and lexical complexity, with the latter often underestimated by ChatGPT. To address these issues, we proposed human-in-the-loop validation, where experts reviewed and refined AI-generated outputs. While artificial intelligence demonstrated efficiency in linguistic quantification of English writing CAF, its variability in nuanced assessments highlights the need for ongoing refinement. Future research should further examine the reliability of artificial intelligence across different writing contexts and analytic measures.
科研通智能强力驱动
Strongly Powered by AbleSci AI