从真实课堂问题出发，研究教师、学习者与技术之间的关系。

以五周线下汉语拼音课程为场景，探讨微课程如何回应近零起点学习者在声调、变调、辨音与实际应用中的困难，并依据学习者反思持续改进设计。Set in a five-week face-to-face Chinese pinyin course, the study explores how micro-learning can address near-beginner learners’ difficulties with tones, tone sandhi, sound discrimination and practical application, while iterating design through learner reflection.

采用 Nation 与 Macalister（2020）课程设计模型，纳入 5 名非汉语母语本科生，通过 5 份结构化开放式反思问卷收集 1,770 字学习反馈。The study applies Nation and Macalister’s (2020) curriculum design model with five non-Chinese-speaking undergraduates and collects 1,770 Chinese characters of feedback through five structured open-ended reflection questionnaires.

使用 Excel 整理资料，结合演绎与归纳编码形成 89 个分析单元与 206 个编码片段，并通过模式编码归纳学习需要、呈现方式、评估反馈与学习灵活性。Data were organised in Excel. Deductive and inductive coding produced 89 units of analysis and 206 coded segments; pattern coding then synthesised learning needs, presentation preferences, assessment feedback and learning flexibility.

学习者最关注实际沟通与声调难点，并偏好真人示范、短视频和视觉标注。设计原则包括循序渐进、真实应用、自主与灵活学习，以及书面和口头评价配合及时反馈。Learners prioritised practical communication and tone difficulties, and preferred live demonstration, short videos and visual annotation. Design principles included sequencing, authentic application, autonomous and flexible learning, and timely written and oral feedback.

负责设计引导学习者开展课后自我反思的指引，并完成反思资料整理、主题编码、分析与研究成果撰写与汇报等工作。I designed the post-class reflection guide, organised and thematically coded the reflections, analysed the evidence, and contributed to the writing and conference presentation of the findings.

论文已被 TechCLE 2026 会议论文集接收、待出版；研究先后在 TechCLE 2026 与第十五届中文教学现代化国际研讨会进行现场汇报。The paper has been accepted for the TechCLE 2026 conference proceedings and is forthcoming. The study was presented at TechCLE 2026 and the 15th International Conference on the Modernization of Chinese Language Education.

作为 Student Helper 参与香港华语作家多语写作研究，为粤语与普通话转换材料的系统整理提供研究支持。As a student helper, I supported research on multilingual writing by Sinophone Hong Kong authors, contributing to the systematic organisation of Cantonese–Putonghua conversion materials.

开展语料采集、粤语—普通话对应整理、分类记录与人工复核，确保资料能够回溯原始文本并保持标注一致。Responsibilities included corpus collection, Cantonese–Putonghua alignment, categorisation and manual verification, ensuring traceability to source texts and consistency of annotation.

收集、整理并校对 2,000+ 条粤语—普通话转换语料，形成结构化语料对照资料。I collected, organised and checked more than 2,000 Cantonese–Putonghua conversion entries, forming a structured parallel dataset.

李碧华《白开水》以香港都市日常与口语化散文呈现本地表达；黄碧云《烈佬传》以粤语书写边缘人物的生命史；董启章《香港字：迟到一百五十年的情书》把香港中文铅字史与虚构叙事交织；黄碧云《温柔与暴烈》以反差鲜明的文字书写香港经验与本地语感。Sources include Li Bihua’s Plain Water, which uses colloquial prose to portray everyday urban Hong Kong; Huang Biyun’s The Fierce Man, a Cantonese-inflected life history of marginal characters; Dung Kai-cheung’s The History of the Adventures of Vivi and Vera, which intertwines Hong Kong typography with fiction; and Huang Biyun’s Gentle and Fierce, whose contrasting prose captures local experience and sensibility.

研究 7 名中高级中文学习者如何借助 ChatGPT 创作中文文化语境剧本，重点分析工具、规则、共同体、分工与教师支持之间出现的矛盾，以及这些矛盾如何被调解。This study examines how seven intermediate-to-advanced Chinese learners used ChatGPT to create cultural-context scripts, focusing on contradictions among tools, rules, community, division of labour and teacher support, and how these contradictions were mediated.

以活动理论为框架，综合分析学习者反思、半结构式访谈、工作坊转录与最终剧本；通过定性编码、矛盾识别与跨来源比较，追踪学习者与生成式人工智能的互动过程。Using activity theory, the study integrates learner reflections, semi-structured interviews, workshop transcripts and final scripts. Qualitative coding, contradiction identification and cross-source comparison trace learner–AI interaction.

区分文化语境提示与程序性提示，并比较学习者将 ChatGPT 视为“助手”或“工具”时的不同实践。结果显示，“先文化语境、后程序步骤”的提示顺序更有助于保障文化适切性。The study distinguishes cultural-context prompts from procedural prompts and compares practices in which learners treated ChatGPT as an assistant or as a tool. A culture-first, procedure-second sequence better supported cultural appropriateness.

研究识别出四类主要矛盾，其中“规则—共同体”矛盾最突出；教师调解的解决率为 81.3%，高于同伴协作的 66.7% 与自我修正的 33.3%。有效路径体现为：规则阐释—教师文化指导—跨文化交际能力提升。Four major contradictions were identified, with rules–community tensions most prominent. Instructor mediation resolved 81.3% of relevant episodes, compared with 66.7% for peer collaboration and 33.3% for self-correction. The effective pathway was rule interpretation, teacher cultural guidance and improved intercultural competence.

负责整体研究设计、数据分析、撰写论文核心部分及整体完善等工作，把课堂活动中的矛盾、调节方式与学习结果组织为可解释的研究证据。I led the overall research design and data analysis, drafted the core sections of the paper and contributed to the full manuscript revision, transforming contradictions, mediation practices and learning outcomes into interpretable evidence.

研究表明，生成式人工智能支持的语言学习并非单纯的人机互动，而是由教师、同伴、规则与文化知识共同调节的社会性学习过程；教师由知识传递者进一步转向矛盾调解者。The study shows that generative-AI-supported language learning is not merely human–computer interaction but a socially mediated process involving teachers, peers, rules and cultural knowledge. The teacher’s role extends from knowledge transmitter to mediator of contradictions.