Scholarlink

This study examines the integration of generative AI (GenAI), specifically ChatGPT, into a virtual learning environment (VLE) aimed at enhancing English communication among Malaysian undergraduate students. The VirtuEnglish system enables users to engage in open-ended conversations with GenAI non-playable characters (NPCs), fostering interactive, scenario-based practice. Seven participants were involved in usability testing and direct observation. The System Usability Scale (SUS) produced an average score of 68.2, with the highest agreement on using the system frequently. Beginner speakers (BS) recorded the highest satisfaction, with SUS scores 72.5. Direct observational data showed differences in pause durations and response times across English proficiency levels, emphasizing the need for real-time scaffolding, particularly for BS. The average task completion rate across three scenarios was 50.79%, with the Restaurant scenario showing the highest completion rate CR in interview floor enquiry (85.71%), order meals (85.71%), self-introduction (85.71%). The lowest CR is company environment enquiry task (28.57%), enquiries about career advancement (0%) were often skipped. GenAI performance demonstrated lowest percentage in input accuracy (75.28%), meaning interpretation accuracy (77.55%), and sentence correctness (82.15%). Key limitations included inaccurate name recognition and difficulty processing disfluent speech. Suggested improvements include adding an undo button, localized content, and adaptive timing features. This research contributes to the development of culturally relevant AI-supported tools to support learners in improving spontaneous English communication.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

VirtuEnglish: Evaluating Generative AI-Driven Virtual Agent for Enhancing English Communication in a Virtual Learning Environment

  • 章节

  • 发表日期：

    2026年

查看全文
  • Chin Hui Ooi,
  • Bee Sian Tan,
  • Soo Keong Ong,
  • Khin Huat Andrew Tan,
  • Thein Lai Wong,
  • P. Sharimila Bai Pandurenga Rao

摘要

This study examines the integration of generative AI (GenAI), specifically ChatGPT, into a virtual learning environment (VLE) aimed at enhancing English communication among Malaysian undergraduate students. The VirtuEnglish system enables users to engage in open-ended conversations with GenAI non-playable characters (NPCs), fostering interactive, scenario-based practice. Seven participants were involved in usability testing and direct observation. The System Usability Scale (SUS) produced an average score of 68.2, with the highest agreement on using the system frequently. Beginner speakers (BS) recorded the highest satisfaction, with SUS scores 72.5. Direct observational data showed differences in pause durations and response times across English proficiency levels, emphasizing the need for real-time scaffolding, particularly for BS. The average task completion rate across three scenarios was 50.79%, with the Restaurant scenario showing the highest completion rate CR in interview floor enquiry (85.71%), order meals (85.71%), self-introduction (85.71%). The lowest CR is company environment enquiry task (28.57%), enquiries about career advancement (0%) were often skipped. GenAI performance demonstrated lowest percentage in input accuracy (75.28%), meaning interpretation accuracy (77.55%), and sentence correctness (82.15%). Key limitations included inaccurate name recognition and difficulty processing disfluent speech. Suggested improvements include adding an undo button, localized content, and adaptive timing features. This research contributes to the development of culturally relevant AI-supported tools to support learners in improving spontaneous English communication.