중국어 학습자 코퍼스 오류 유형 분석과 자동 피드백 모델 연구 — ChatGPT와 Claude를 중심으로

초록

This study aims to propose a theoretical and technological framework for transforming error annotation in Chinese learner corpora into an educational feedback system. Drawing on the multi-layered annotation principles of the HSK Dynamic Composition Corpus, the study restructures learner errors into lexical and grammatical domains and develops an integrated model encompassing automatic annotation, correction, and explanation. Methodologically, the framework incorporates the generalization capacity of large language models while explicitly organizing the reasoning flow of “grounding–adaptation–explanation” through a triple mechanism of Retrieval-Augmented Generation (RAG), In-Context Learning (ICL), and Chain-of-Thought (CoT). The proposed model demonstrates broad pedagogical appstic analysis and classroom feedback practice. As a result, this framework substantiates the pedagogical potential of AI-assisted automatic error annotation, bridging the analylicability of error annotation, enabling a dynamic connection between corpus-based linguitical rigor of corpus linguistics with the practical efficiency of Chinese language education.

키워드