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

Langova functions as a WebGL-based serious game designed to bridge the gap between theoretical English language acquisition and practical, real-world communication. By simulating common public scenarios such as airports and restaurants, the platform provides an immersive, low-pressure environment that allows ESL learners to practice conversational skills. The primary thesis posits that integrating speech recognition technology within a structured, interactive game environment effectively reduces language anxiety, thereby fostering greater fluency and confidence compared to traditional, passive educational tools.

The platform utilizes a three-act structure and adaptive similarity thresholds to provide immediate, context-aware feedback. Empirical testing involving 57 university students confirms the system's efficacy, with performance quantified through Word Error Rate (WER) and Average Word Error Rate per Sentence metrics, which yielded consistent results between 0.23 and 1.22. High transaction volumes observed during these trials served as a key indicator of user engagement, revealing that learners often utilized the platform for extensive repetition to refine their pronunciation and overcome speaking-related barriers.

While the research demonstrates significant potential for serious games in language pedagogy, technical limitations remain a factor in overall performance. Factors such as background noise, microphone quality, and regional accent variations occasionally impede speech recognition accuracy. Future development efforts are directed toward integrating more advanced, inclusive speech recognition models and adaptive learning features to enhance the system's robustness. Ultimately, the study establishes a strong technical and pedagogical foundation for utilizing interactive, scenario-based digital environments to support English language learners in developing practical communication skills.