A WebGL Serious Game for Practicing English Conversations in Public Places Using Speech Recognition

Abstract

Learning English as a Second Language (ESL), particularly for conversation, often presents challenges, especially due to learners' fear of judgment during public practice. Current tools lack the ability to provide private, practical, and interactive speaking environments. This study investigates how to create a private, practical, and interactive platform for English conversation practice. To address this gap, Langova, a WebGL-based online game leveraging Speech Recognition technology, was developed to facilitate English conversation practice in realistic public scenarios, such as airports, hotels, and malls. This approach enables learners to practice privately while building confidence. The study involved 57 university-level ESL students with varying levels of English proficiency, who participated in a three-month trial of Langova. The methodology includes system analysis using Word Error Rate (WER) and Average Word Error Rate per Sentence (Avg WER) to evaluate speech recognition accuracy. A trial with 57 students revealed WER values ranging from 0.23 to 1.22 and Avg WER values from 0.15 to 1.99, highlighting variability and feedback consistency. Compared to traditional learning methods, Langova provides an innovative, accessible, and interactive platform for language acquisition. By combining speech recognition technology with immersive learning environments, Langova empowers learners to build confidence and improve fluency. The platform also provides actionable data for educators to design targeted interventions, suggesting broader implications for integrating similar technologies into language education frameworks. Qualitative feedback from participants indicated that Langova’s private and flexible environment can reduce their anxiety about speaking English, while also revealing technical challenges such as microphone sensitivity and accent recognition.