Fostering English Speaking Proficiency through Deep Learning-Based Habitual Speaking Practices (A Study at SMP Negeri 24 Semarang)

Abstract

The ability to communicate effectively in English is an important goal of English language education, yet junior high school learners may have limited opportunities to use English orally in meaningful classroom contexts. This study explores the implementation of deep learning-based habitual speaking practices at SMP Negeri 24 Semarang. The practice was grounded in Mindful Learning, Meaningful Learning, and Joyful Learning and supported by game-based activities and AI-based learning tools. A qualitative approach was employed using classroom observations, student questionnaires, and teacher questionnaires. Thirty-five students and three English teachers participated in the evaluation. Student responses across 25 Likert-scale statements produced an overall mean of 4.58/5, while teacher responses across 30 statements produced an overall mean of 4.67/5. Students particularly reported that the activities helped them use transactional expressions, understand buyer/seller roles, and practise speaking through market simulation. Teacher responses consistently rated the learning design, deep learning characteristics, student engagement, and applicability positively. Qualitative responses further identified market simulation as the most frequently mentioned preferred activity and highlighted increased confidence, practical communication, collaboration, and enjoyment. The findings indicate that habitual speaking supported by contextual games and simulation can foster active, meaningful, and enjoyable English speaking practice.