VerbVenture──混合實境與人工智慧，輔助兒童英語搭配詞學習之設計與評估

VerbVenture: Designing and Evaluating a Mixed Reality and AI-Supported System for Children’s English Collocation Learning

本研究旨在驗證一套應用於兒童英語搭配詞（collocation）學習的混合實境（MR）教學設計。此設計基於具身認知理論，旨在將日常環境轉化為沉浸式語言學習場域，並用以回應EFL兒童因母語影響及缺乏練習情境所面臨的挑戰。為在理想回饋條件下評估此教學設計的有效性，本研究開發了原型VerbVenture，其學習流程分為兩階段：（1）詞彙習得：透過人工智慧（artificial intelligence, AI）物件辨識，學習者能探索真實物件的英文單字與發音；（2）搭配詞練習：系統播放情境動畫，並以動態提示引導學習者操作實體或虛擬物件演練動詞－名詞等搭配詞用法，從而強化語言與身體動作的連結。本研究對14位國小學童進行了前測—後測—延後測的受測者內實驗，比較操作實體與虛擬物件的學習差異。量化結果顯示，無論操作何種物件，VerbVenture系統皆能帶來顯著的學習進步與記憶保留效果。雖然兩種互動模式在學習成效上未呈現統計上的顯著差異，但質性分析揭示了其互補價值：實體物件提供真實的觸覺回饋與物理特性，讓學習者感到更直觀、易於掌控，有助於建立初期的動作記憶。虛擬物件憑藉超越現實物理限制的新奇感，顯著提升了學習者的參與動機與想像力，且可能有助於長期記憶保留。本研究的貢獻在於，在理想互動條件下驗證了MR具身學習教學設計的成效，並基於實證歸納出一套以具身學習為核心的MR學習設計原則，為未來開發更具吸引力與成效的智能化語言學習工具提供具體的設計指引。

This study validates an innovative mixed reality (MR) instructional design for children’s English collocation learning. Grounded in embodied cognition theory, the design aims to transform everyday environments into immersive language learning spaces, addressing challenges EFL learners face due to first-language interference and limited contextual practice. A prototype system, VerbVenture, was developed to implement this approach through two learning stages: (1) Vocabulary Acquisition—AI-based object recognition helps learners explore real-world objects and learn their English names and pronunciations; and (2) Collocation Practice—contextual animations and dynamic prompts guide learners to manipulate physical or virtual objects, enacting verb–noun collocations to strengthen language-action connections.
A within-subjects experiment with 14 elementary students compared physical and virtual object interactions using pre-, post-, and delayed post-tests. Both modes significantly improved learning and retention, though no statistical difference emerged between them. Qualitative findings revealed complementary advantages: physical interactions fostered intuitive control and motor memory, while virtual ones enhanced engagement and imagination.
This study contributes empirical evidence for MR-based embodied learning and derives design principles for integrating embodied cognition and AI technologies (e.g., object, motion, and speech recognition) in educational MR systems, informing the design of engaging, intelligent language learning tools.