Download Full PDF

周万勤

朱宇 

厦门大学，中国

 

摘要

随着大语言模型与语料库技术的发展，专门用途汉语教学资源的动态生成逐渐成为可能。本文以中医汉语教学为例，基于自建专业语料库并结合 RAG 技术，构建了一套支持知识总结、问答设计、文化分析与教学计划自动化的资源生成框架。该框架以语料库为核心，充分利用多项语料库来约束和优化生成过程，体现了基于语料库的语言教学以真实语言证据支撑教学设计的原则。对比双盲打分结果显示，语料库驱动生成在教学适切性方面显著优于无语料库驱动。教师访谈进一步表明，该方法能够有效提升备课效率，提升资源可用性与可信度，成为强实用性的教学平台。未来，可通过扩展专门用途汉语语料库，并结合大语言模型与多模态方法，生成更高质量的垂直领域教学资源。

 

关键词

语料库，教学资源生成，语言模型，专门用途汉语

 

Application Research on Corpus-Enhanced RAG Large Language Models in the Generation of Teaching Resources for Chinese as a Foreign Language in Traditional Chinese Medicine
 

Wanqin Zhou

Yu Zhu

Xiamen University, China

 

Abstract

With the advancement of Large Language Models (LLMs) and corpus technologies, the dynamic generation of teaching resources for Languages for Specific Purposes (LSP) has gradually become feasible. Taking the teaching of Chinese for Traditional Chinese Medicine (TCM) as a case study, this paper constructs a resource generation framework supporting knowledge summarization, question-and-answer design, cultural analysis, and automated instructional planning. This framework is built upon a self-constructed specialized corpus and integrates RetrievalAugmented Generation (RAG) technology. Centered on the corpus, the framework leverages multiple corpora to constrain and optimize the generation process, embodying the principle of corpus-based language pedagogy where instructional design is underpinned by authentic linguistic evidence. Results from a double-blind scoring comparison indicate that corpus-driven generation significantly outperforms non-corpus-driven methods in terms of pedagogical appropriateness. Furthermore, teacher interviews reveal that this approach effectively enhances lesson preparation efficiency, improves the usability and credibility of resources, and serves as a highly practical teaching platform. Future research may focus on expanding LSP Chinese corpora and integrating LLMs with multimodal methods to generate higher-quality teaching resources for vertical domains.

 

Keywords

Corpus, teaching resource generation, language models, Chinese for specific purposes