リングローマ (Ringurooma) - A website for evaluating Japanese speaking skills

P/s: This project is using its own custom speech service at here

Introduction

Ringurooma is a web-based platform designed to help users improve their Japanese speaking skills using AI-driven scoring and personalized feedback. Powered by Azure AI, the system evaluates pronunciation, fluency, and prosody, then delivers constructive suggestions tailored to each learner.

Key Features

• Prompt-based Speaking Practice: Choose a topic, speak, and receive instant feedback.

• Accurate Pronunciation Scoring: Analyze accuracy, fluency, and prosody, and estimate JLPT level.

• AI Feedback: Interactive AI chatbot gives scores, strength/weakness analysis, and improvement tips.

• Personalized Experience: Customize chatbot tone and learning style based on user profile.

• Shadowing and Intent Recognition Support: Practice native-like speech and let AI understand your learning intent.

Problems Solved

• Most Japanese learners in Vietnam struggle with speaking due to lack of effective tools.

• 80% of existing apps (e.g., Duolingo, Memrise) focus only on vocabulary/grammar, while HelloTalk lacks personalization.

• Ringurooma bridges this gap by combining interaction with AI-based assessment and feedback.

Architecture Decisions & System Decomposition

• Hybrid architecture enables separation of frontend/backend for scalable development

• Low-code platforms (Appsmith, n8n) enable fast iteration with minimal technical overhead

• Ringurooma Speech Service acts as a reusable, standalone API layer for speech assessment

System Architecture

• Frontend: Appsmith (low-code)

• Backend: n8n (workflow automation)

• AI Service: Ringurooma Speech Service (custom proxy to Azure AI)

• Cloud AI: Azure Speech & Language Services

• Database: PostgreSQL

• Architecture: Hybrid (Client-Server + Service-Oriented)

• Languages: HTML, CSS, JavaScript, Python

Database Design

See the full Database Design for detailed schema, tables, and relationships used in Ringurooma.

Demo Video

Watch a full walkthrough of Ringurooma in action:

Casual Chatting

Link video

Topic Suggestion

Link video

Speech Suggestion

Link video

Speech Prompt Suggestion

Link video

Speaking Evaluation

Link video

Testing

Cypress is used for automated testing of UI flows including login, signup, chat interaction a. n8n automation test is used for backend logic testing, ensuring workflows execute correctly, calculations are accurate, and database interactions behave as expected.

References

• Statista, JLPTSensei, Vietnam Market Research

• Microsoft Azure AI SDK

• Detailed Documentation (PDF)

Future Development Plans

Here are some planned directions to expand Ringurooma's capabilities in the future:

• Mobile App Version: Build a native mobile version using React Native or Flutter to enhance accessibility.

• Real-time Voice Chat with Native Tutors: Integrate real-time speaking sessions with native Japanese speakers.

• Multi-language Support: Expand the platform to support other languages (e.g., Korean, English).

• Advanced AI Scoring Models: Train custom ML models to improve feedback precision beyond Azure.

• Gamification Features: Add streaks, XP points, badges, and challenges to boost user motivation.

• A/B Testing Framework: Integrate testing tools to continuously optimize UI/UX and feedback style.

• Admin Dashboard: Provide data visualization and learner performance tracking for institutions.