Scholarlink

Collocations, frequently occurring word combinations, are crucial for achieving native-like fluency in English, yet their acquisition remains a significant challenge for EFL (English as a Foreign Language) students. While existing CALL (Computer-assisted language learning) tools address collocation learning and pronunciation practice separately, a critical gap persists in integrated solutions. We introduce ColloSpeak, a GenAI-powered web system designed to bridge this gap by combining collocation learning with pronunciation-focused speaking practice. ColloSpeak offers reliable collocation search outputs, mitigating AI hallucination, and incorporates speech recognition for immediate, personalized feedback on pronunciation. It also facilitates practical speaking exercises based on learned collocations, leveraging a cloud-based speech service and a large language model. An evaluation involving 28 Japanese English learners (beginner to advanced) revealed positive reception of the system in terms of learner fit, meaning-focused learning, and overall experience. Furthermore, an assessment through Chappelle’s CALL task framework confirmed ColloSpeak’s pedagogical soundness. Our findings suggest that AI-based CALL tools like ColloSpeak can significantly enhance collocation-centered language learning and promote second language acquisition in a more efficient and engaging manner.

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

Large Language Models for Improving Speaking Proficiency in EFL Contexts: A Collocation-Centric Approach

  • 章节

  • 发表日期：

    2026年

查看全文
  • Emi Ichimura,
  • M. Fahim Ferdous Khan,
  • Ken Sakamura

摘要

Collocations, frequently occurring word combinations, are crucial for achieving native-like fluency in English, yet their acquisition remains a significant challenge for EFL (English as a Foreign Language) students. While existing CALL (Computer-assisted language learning) tools address collocation learning and pronunciation practice separately, a critical gap persists in integrated solutions. We introduce ColloSpeak, a GenAI-powered web system designed to bridge this gap by combining collocation learning with pronunciation-focused speaking practice. ColloSpeak offers reliable collocation search outputs, mitigating AI hallucination, and incorporates speech recognition for immediate, personalized feedback on pronunciation. It also facilitates practical speaking exercises based on learned collocations, leveraging a cloud-based speech service and a large language model. An evaluation involving 28 Japanese English learners (beginner to advanced) revealed positive reception of the system in terms of learner fit, meaning-focused learning, and overall experience. Furthermore, an assessment through Chappelle’s CALL task framework confirmed ColloSpeak’s pedagogical soundness. Our findings suggest that AI-based CALL tools like ColloSpeak can significantly enhance collocation-centered language learning and promote second language acquisition in a more efficient and engaging manner.