大语言模型在语言测评中的应用  

Application of large language models in language testing and assessment

作　　者：何莲珍[1] HE Lianzhen(School of Intermational Studies,Zhejiang University,Hangzhou 310058,China)

机构地区：[1]浙江大学外国语学院,浙江省杭州市310058

出　　处：《外语教学与研究》2024年第6期903-912,960,共11页Foreign Language Teaching and Research

摘　　要：本文从语言测评的研发、实施、评分与反馈三方面梳理大语言模型在语言测评领域的应用前景,指出大语言模型应用于语言测评所面临的四大挑战,即大语言模型的“幻觉”问题、不可解释性问题、概化问题和相关伦理问题。本文提出,语言测评领域应主动迎接挑战,从理论和实践两方面对语言测评的效度以及测评结果的解释和使用进行深入研究,推动语言测评的内涵式高质量发展,确保教育公平。本文同时呼吁语言测评与人工智能领域专家学者应深入交流,密切合作,实现共赢。This article reviews and discusses the prospects of using large language models in language testing and assessment,focusing on three key stages:test development,test administration,and rating and feedback.Four major challenges in the field of language testing and assessment are highlighted:the"hallucination"problem,the issue of uninterpretability,of generalizability,and of relevant ethical concerns.This article proposes that to promote the high-quality and sustainable development of language testing and assessment to ensure fairness in education,language testing and assessment professionals should embrace the challenges,conduct in-depth theoretical and empirical research,especially on the validity of testing and assessment and the interpretation and use of their results.In addition,experts and scholars in the fields of language testing and assessment and artificial intelligence should engage in deep exchanges and close cooperation for mutual benefits.

关 键 词：大语言模型 语言测评 应用前景 现实挑战 

分 类 号：H319.6[语言文字—英语]