TUKI TOEIC: An AI-Powered Personalized TOEIC Preparation System

Các tác giả

  • Tú Nguyễn Tâm Trường Đại học Công nghiệp TP.HCM
  • Vy Trần Thị Thúy
  • Khánh Nguyễn Thị Hoàng

Từ khóa:

Cosine Similarity, Natural Language Processing, TF-IDF, Vector Embedding., Personalized Learning, Speaking Assistant

Tóm tắt

The TUKI TOEIC system is developed to support English learners in preparing for the TOEIC examination with a personalized and adaptive learning pathway. The platform provides practice for all four skills (Listening, Reading, Speaking, and Writing), with a particular focus on interactive speaking training through the Tuki Speaking Bot, an AI-based conver sational partner that offers instant feedback on pronunciation, grammar, and fluency. TUKI TOEIC is implemented on both web and mobile platforms using Next.js, NestJS, and React Native, combined with modern AI technologies for speech recognition, text-to-speech, and natural language processing. Personalization in TUKI TOEIC is driven by semantic embedding models and cosine similarity, where each learner’s utterance or response is converted into a semantic vector. This enables the system to provide appropriate feedback and content for each user, while understanding their strengths and areas for improvement to optimize the conversational experience. In addition, the system employs a personalized learning-path algorithm, spaced repetition-based vocabulary tracking, and a Retrieval-Augmented
Generation (RAG) speaking bot integrated with Qdrant and Gemini. Together, these components turn TUKI TOEIC from a traditional exam-preparation tool into an intelligent adaptive learning platform that continuously monitors progress, provides targeted feedback, and helps learners improve their TOEIC skills efficiently and naturally.

Đã Xuất bản

09-12-2025

Số

Chuyên mục

Kỹ thuật phần mềm (Software Engineering)