<![CDATA[This study investigated the efficacy of Artificial Intelligence (AI)-based Text-to-Speech (TTS) technology in Indonesian pronunciation training for Foreign Speakers (BIPA), addressing a notable research gap concerning its specific effectiveness and learner perceptions. Employing a mixed-method quasi-experimental design with experimental (n=20) and control (n=20) groups, the research utilized pronunciation tests, perception questionnaires, and interviews. Based on the paired sample t-test, findings showed that AI TTS was significantly more effective than conventional methods in improving BIPA learners’ pronunciation in accuracy, fluency, and intelligibility. This efficacy is attributed to AI's capacity for immediate, personalized feedback and objective analysis. The qualitative data analysis revealed that learners reported overwhelmingly positive perceptions regarding AI TTS's effectiveness, engagement, and confidence-boosting impact, appreciating its non-judgmental and accessible nature. However, limitations emerged concerning voice naturalness, intonation accuracy, and the interpretation of contextual nuances. Concerns were also raised about potential over-reliance on AI, technical reliability, and data privacy. These findings strongly advocate for a blended learning approach in BIPA pronunciation instruction, strategically leveraging AI's strengths while preserving the essential value of human teaching for higher-order linguistic and cultural competence. The study contributes to applied linguistics by providing empirical insights into AI applications in second language acquisition beyond English contexts and offers practical guidance for developing adaptive, user-centered BIPA curricula and fostering responsible AI integration.]]>
Copyrights © 2025
