Scholarlink

Generative AI is revolutionizing education by providing personalized and interactive learning experiences. In primary education, AI-powered systems can significantly enhance language acquisition through simulated conversations and instant feedback. However, challenges exist in ensuring scalability, contextual relevance, and the appropriateness of AI-generated content for young learners. This chapter presents a scalable AI-assisted platform designed to improve oral proficiency in Mother Tongue Languages (MTL), notably Chinese, Tamil, and Malay, for Primary 1 and 2 students as part of Singapore national curriculum. The platform leverages a modular architecture to facilitate realistic, engaging, and coherent short conversations between students and an AI agent in their MTL, aiming to improve oral proficiency. The platform facilitates two main use cases: (1) for teachers to generate AI-assisted questions based on images, and (2) for students to engage in structured, interactive conversations with the AI in their MTL. It integrates four core AI components: the Visual Question & Answering Engine (VQA) for generating contextually relevant questions, the Translation Manager (TM) for accurate multilingual translations, the Dialogue Manager (DM) for managing AI-assisted conversations, and the Speech Engine (SE) for real-time Text-to-Speech (TTS) and Automatic Speech Recognition (ASR). The system’s frontend and backend architecture is designed for scalability and efficiency, ensuring reliable performance for diverse user groups. The AI platform was deployed in three phases: (1) multiple-choice questions (MCQ) to assess TTS capabilities, (2) a word bank exercise for ASR evaluation, and (3) a conversational chatbot to test the full range of AI capabilities under real-world conditions. A pilot study involving 206 students across different MTL groups showed promising results, with high levels of engagement and positive feedback on the interactive tasks. The findings suggest that AI-assisted learning can significantly enhance oral proficiency and engagement in MTL learning. In summary, the proposed system presents a practical and scalable solution for MTL education, addressing both linguistic diversity and resource constraints. Future research will focus on validating learning outcomes, enhancing gamification elements, and refining cultural adaptations to further advance the capabilities of AI-assisted multilingual education.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Leveraging Generative AI for Primary School Mother Tongue Language Learning: A Scalable and Adaptive AI-Driven Approach

  • 章节

  • 发表日期：

    2026年

查看全文
  • Donny Soh,
  • Indriyati Atmosukarto,
  • Tong Rong,
  • Ian McLoughlin,
  • Lim Jing Hao,
  • Mohamed Sirajudeen,
  • Justin Kan,
  • Pranav Tushar,
  • Zhang Bowen,
  • Nur Afiqah Bte Abdul Latiff,
  • Loo Yin Yin

摘要

Generative AI is revolutionizing education by providing personalized and interactive learning experiences. In primary education, AI-powered systems can significantly enhance language acquisition through simulated conversations and instant feedback. However, challenges exist in ensuring scalability, contextual relevance, and the appropriateness of AI-generated content for young learners. This chapter presents a scalable AI-assisted platform designed to improve oral proficiency in Mother Tongue Languages (MTL), notably Chinese, Tamil, and Malay, for Primary 1 and 2 students as part of Singapore national curriculum. The platform leverages a modular architecture to facilitate realistic, engaging, and coherent short conversations between students and an AI agent in their MTL, aiming to improve oral proficiency. The platform facilitates two main use cases: (1) for teachers to generate AI-assisted questions based on images, and (2) for students to engage in structured, interactive conversations with the AI in their MTL. It integrates four core AI components: the Visual Question & Answering Engine (VQA) for generating contextually relevant questions, the Translation Manager (TM) for accurate multilingual translations, the Dialogue Manager (DM) for managing AI-assisted conversations, and the Speech Engine (SE) for real-time Text-to-Speech (TTS) and Automatic Speech Recognition (ASR). The system’s frontend and backend architecture is designed for scalability and efficiency, ensuring reliable performance for diverse user groups. The AI platform was deployed in three phases: (1) multiple-choice questions (MCQ) to assess TTS capabilities, (2) a word bank exercise for ASR evaluation, and (3) a conversational chatbot to test the full range of AI capabilities under real-world conditions. A pilot study involving 206 students across different MTL groups showed promising results, with high levels of engagement and positive feedback on the interactive tasks. The findings suggest that AI-assisted learning can significantly enhance oral proficiency and engagement in MTL learning. In summary, the proposed system presents a practical and scalable solution for MTL education, addressing both linguistic diversity and resource constraints. Future research will focus on validating learning outcomes, enhancing gamification elements, and refining cultural adaptations to further advance the capabilities of AI-assisted multilingual education.