Anonymized Dataset and Scripts for LLM Intervention in Chinese EFL: Reducing FLA and Enhancing Proficiency

Anonymized dataset and analysis scripts supporting "Large Language Models, Speaking Anxiety, and Self-Efficacy in Chinese EFL." This dataset covers an 8-week LLM-based speaking intervention (October 22 - December 17, 2025) conducted with 54 intermediate-level Chinese EFL students at a university in China, using DeepSeek, Doubao, and Kimi Chat (among other tools) for low-stakes speaking practice delivered via WeChat. Contents: pre/post scores on the FLCAS (Foreign Language Classroom Anxiety Scale), ESSES (English Speaking Self-Efficacy Scale), VRT (Voice Recording Task proficiency rubric), and OSE (Online Student Engagement Scale); anonymized transcripts and thematic coding from 12 semi-structured interviews (437 speaker turns); weekly AI-practice engagement logs; and the Python scripts used to process and score all quantitative measures. Version note: this release corrects two issues present in an earlier internal version: a duplicated participant record in the master quantitative file and an item-scoring bug that omitted one FLCAS item from computed totals. See README.txt for full details. Citation: Please cite the associated paper and this dataset DOI if you use this data. Re-users must not attempt to re-identify participants.