DASI English (다시 영어) is an intelligent platform designed to guide users from basic vocabulary to native-level fluency through a structured, data-driven learning path that mirrors natural language acquisition.
Our core philosophy is built on two pillars:
- Data-Driven Vocabulary Stratification: We use word frequency analysis to teach the most effective vocabulary first, ensuring every lesson provides maximum communicative power.
- Simulated Growth: The 10-level system simulates the human journey of learning a language—from single-word utterances as a "Newborn" to expressing nuanced, idiomatic thoughts as an "Expert."
This project is not just a conversation partner; it is a sophisticated ecosystem that intelligently manages learning, retention, and motivation.
Our system is more than just a series of levels; it's a carefully designed growth path. Each stage equips the user with specific tools to unlock the next phase of communication, with the AI's persona and conversational style evolving alongside the learner.
| Level | Title | Evolutionary Leap | Example Interaction |
|---|---|---|---|
| 1-2 | Newborn / Toddler | From words to simple sentences. | User: "Apple." → AI: "Yes, this is an apple. Do you like apples?" |
| 3-4 | Child | From sentences to simple reasoning. | User: "I like it." → AI: "That's great! Why do you like it?" |
| 5-6 | Teenager | From reasoning to structured arguments. | User: "It's good." → AI: "I see. But although it's good, it's also expensive. What do you think?" |
| 7-8 | University Student | From arguments to abstract thought. | User: "We need a new policy." → AI: "Interesting. What is your take on the potential social impact of such a policy?" |
| 9-10 | Professional / Expert | From fluency to nuanced, culturally-aware communication. | User: "It's a good solution." → AI: "I agree. But it could be a double-edged sword, so to speak." |
For a detailed breakdown of the curriculum, AI personas, and learning mechanics at each level, please see our Learning System Design.
- Adaptive AI Conversation: Utilizes a hybrid of GPT-4o and Claude 3.5 Sonnet to provide conversations that are perfectly matched to the user's proficiency level.
- High-Fidelity Speech & Pronunciation Analysis: Powered by Whisper Large v3 for accurate transcription and detailed phonetic feedback.
- Forgetting Curve Engine: A custom-built system (Python + Firebase) that schedules personalized review sessions to ensure long-term retention.
- Gamified Progression: An EXP and streak system designed to motivate consistent practice.
- **Intelligent Feedback:**0 Real-time, level-appropriate corrections, nuance explanations, and paraphrasing suggestions.
The system integrates a suite of best-in-class AI models and technologies to create a seamless learning loop. For a complete overview of the system design, AI model roles, and data flow, please refer to our System Architecture Document.
- Frontend: React (TypeScript)
- Backend: Node.js (Express), Python
- AI Models: OpenAI GPT-4o, Anthropic Claude 3.5, OpenAI Whisper v3, ElevenLabs TTS
- Database: Firebase Firestore
We are developing DASI English following a phased, agile approach focused on rapid iteration and data-driven decisions. To see our strategic plan, from initial technical validation to ecosystem expansion, please view our Development Roadmap.
Current Implementation Status (2025-01-12):
- ✅ Level 1: 376 core expressions (Pattern Viewer)
- ✅ Level 2: 20 stages, Speaking-based learning
- ✅ Level 3: 30 stages, 6 phases (currently Writing-based)
- 🔄 In Progress: Converting Level 3 to Speaking-based for consistency
Speaking vs Writing Modes:
- Primary Focus: Speaking-based learning (Korean audio → English speech)
- Secondary Feature: Writing mode available as backup/alternative
- Current Priority: Unified Speaking experience across all levels
Completed UI Improvements:
- ✅ Grid layout for Level 2 (20 stages → 2x10 grid)
- ✅ Grid layout for Level 3 (30 stages → 3x10 grid)
- ✅ AI coaching system for both levels
- ✅ Phase indicator system for Level 3
Backup & Archive:
- 📁 Writing-based Level 3 backed up in
/backup/directory - 🎨 UI improvements documented and preserved
- 🔄 Ready for future Writing mode implementation
/
├── backend/ # Node.js backend server & Python services
├── web_app/ # React frontend application
├── flutter_app/ # (Future) Flutter mobile application
└── docs/ # Detailed documentation
├── ARCHITECTURE.md
├── DEVELOPMENT_ROADMAP.md
└── LEARNING_SYSTEM.md