How thiagohgl/ai-pronunciation-trainer works
Main Components to Include:
Frontend Layer
• Static assets (CSS/JavaScript) components
Backend Web Server
• WebApp.py as the main server component
• Use green for server components
Core Processing Components
• Speech Processing Module (whisper_wrapper.py)
• AI Models Module (AIModels.py)
• Rule-Based Models Module (RuleBasedModels.py)
• Word Processing Modules (WordMatching.py, WordMetrics.py)
• Use orange for processing components
Lambda Functions Layer
• Sample Generation (lambdaGetSample.py)
• Speech Scoring (lambdaSpeechToScore.py)
• Text-to-Speech (lambdaTTS.py)
• Use purple for lambda components
Data Storage
• Databases component (databases folder with CSV files)
• Pickle file storage (data_de_en_2.pickle)
• Use yellow for data storage components
Connection Flows:
User Interaction Flow
→ Frontend to web server
→ Web server to processing components
Processing Flow
→ Connect lambda functions to their respective processing components
→ Show data flow between storage and processing components
Specific Implementation Guidelines:
[User Browser] → [Frontend (HTML/JS/CSS)] → [Web Server (Flask)]
• ↑ ↓
• [Processing Pipeline]
• ↓
• [Data Storage]
[Speech Input] → [Whisper ASR] → [AI Models]
• ↓
[Database] ←→ [Word Matching] ← [Rule-Based Models]
• ↓
• [Score Output]
Key Architectural Patterns to Highlight:
Model-View-Controller (MVC) pattern
• View: templates/static files
• Controller: webApp.py
• Model: AIModels.py, RuleBasedModels.py
Interface-based design
• ModelInterfaces.py as a connection point
• Extensibility for new languages
Service-oriented architecture
• Lambda functions as independent services
• Clear separation of concerns
• Use rectangles for standard components
• Use hexagons for server components
• Use clouds for external services (like ffmpeg)
• Notes about language extension points
• Document external dependencies
Level of Detail:
• Clear separation between web interface, processing logic, and data storage
• Extension points for new languages as mentioned in the README
• ASR model integration points
Remember to add notes about:
• FFmpeg dependency
• Browser compatibility (Chrome-focused)
• Language extension points
• Model interfaces for new ASR implementations