生成式人工智能赋能国际中文教学的效果测评——以教学设计、HSK模拟试题编写及作文评分为例  

Measuring the Effectiveness of Generative Artificial Intelligence in Enabling International Chinese Language Teaching:A Case Study of Teaching Content Design,HSK Practice Test Writing,and Essay Scoring

作　　者：刘玉屏[1] 欧志刚 武晓琴 LIU Yu-ping;OU Zhi-gang;WU Xiao-qin(College of International Education,Minzu University of China,Beijing 100081)

机构地区：[1]中央民族大学国际教育学院,北京100081

出　　处：《民族教育研究》2025年第1期156-166,共11页Journal of Research on Education for Ethnic Minorities

基　　金：2024年度国家社科基金项目“面向国际中文教育的语法成果教学转化资源智慧平台建设”(项目编号:24BYY046)的阶段性成果;世界汉语教学学会全球中文教育主题学术活动计划项目“AIGC国际中文多模态教学资源的生成与评估研究”(项目编号:SH23Y33)的阶段性成果。

摘　　要：生成式人工智能可辅助教师完成多种教学活动,为教师减负增效,但生成过程需要提示文本指导,生成结果需要评估,技术模型需要对比选择。鉴于此,本文以生成式人工智能生成教学设计、编写HSK模拟试题、对作文进行评分三项任务为例,编写提示文本驱动大语言模型生成内容,设计评价方法和流程对生成内容进行测评,并基于测评结果对不同大语言模型的生成能力进行评述。研究发现:基于框架生成的提示文本可以驱动大语言模型顺利完成任务,参与实验的三个大语言模型生成能力各具特色,均取得了较好的评价结果。研究也证实生成式人工智能生成的内容并不完全符合预期,仍需人工把控,建议在大语言模型实践中优化技术选择和开展技术培训,建立内容审核与动态调整机制,探索融合创新的教学模式以及加强隐私保护与伦理规范。Generative AI can assist teachers in accomplishing a variety of teaching activities and reduce the burden and increase the efficiency for teachers,but the generation process requires prompt textual guidance,the generation results need to be evaluated,and the technology models need to be compared and selected.In view of this,this paper takes the three tasks of generative artificial intelligence to generate instructional design,write HSK practice test questions,and grade essays as examples,writes prompts for text-driven large language model-generated content,designs evaluation methods and processes for measuring the generated content,and reviews the generative ability of different large language models based on the evaluation results.The study found that the framework-generated prompt text can drive the large language model to complete the tasks successfully,and the three large language models participating in the experiment have their own distinctive generative capabilities,all achieving better evaluation results.The study also confirms that the AI-generated content does not meet the expectations absolutely and still needs to be controlled manually,and recommends optimizing the technology selection and technical training in the practice of large language modeling,establishing a mechanism for content review and dynamic adjustment,exploring the integration of innovative teaching modes,and strengthening the privacy protection and ethical norms.

关 键 词：生成式人工智能 大语言模型 国际中文教学 赋能效果测评 

分 类 号：H195[语言文字—汉语] G434[文化科学—教育学]