🇧🇩 Banglish-to-Bangla: Phonetic Smart Converter

An intelligent, rule-based, and dictionary-validated NLP engine designed to convert Romanized Bengali (Banglish) into authentic Bangla Unicode script. Features dynamic vowel-expansion for informal texting/chat shorthand, weighted multi-dictionary validation across 464,000+ words, and an alphabetically sorted curated base dictionary of 2,770+ words covering 80%+ of daily conversational Bengali.

🚀 Key Features

• Dynamic Vowel-Expansion & Shorthand Recovery: Seamlessly transliterates informal chat shorthand (e.g., tmi → তুমি, vlo → ভালো, kmn → কেমন, amr → আমার, bndhu → বন্ধু, rsta → রাস্তা).
• Curated 2,770+ Baseline Dictionary: Pre-indexed and alphabetically sorted mapping in data/ben2bn.csv covering top 80%+ everyday Bengali vocabulary.
• Weighted Multi-Tier Dictionary: Validates candidates against 464,411 words across 4 frequency tiers for maximum accuracy.
• Modern Glassmorphic Web UI: Responsive web app with live transliteration, history, voice typing, and clipboard utilities.
• FastAPI Backend: Clean RESTful endpoints (/convert, /stats) for integration into web, mobile, and desktop applications.
• In-Memory Cache: High-speed session caching (b2b_cache) without disk pollution.
• PWA & Offline Ready: Service Worker and Web Manifest support for mobile/desktop installability.

💻 Installation & Quick Start

1. Clone the Repository

git clone https://github.com/skhalidmahmud/ShobdoSearch.git
cd ShobdoSearch

2. Start the Web App (Recommended)

This installs required dependencies and launches the local dev server:

python run.py

Open your browser at http://localhost:8080.

3. Using Docker

docker build -t banglish-converter .
docker run -p 8080:8080 banglish-converter

4. Interactive CLI Mode

python converter.py

🧪 Testing & Verification

Run the built-in test suite to verify phonetic accuracy, edge cases, and chat shorthand recovery:

python verify.py

Output:

===========================================
      SHOBDOSEARCH CONVERSION TESTS       
===========================================
 [PASS] ka           -> কা
 [PASS] kha          -> খা
 [PASS] eta          -> এটা
 [PASS] kemon        -> কেমন
 [PASS] ami          -> আমি
 [PASS] tumi         -> তুমি
 [PASS] valo         -> ভালো
 [PASS] shundor      -> সুন্দর
 [PASS] manush       -> মানুষ
 [PASS] ghor         -> ঘর
 ...
-------------------------------------------
Total: 41 | Passed: 41 | Failed: 0
 ALL TESTS PASSED SUCCESSFULLY!

🛠️ API Reference

• Interactive Documentation: http://localhost:8080/docs
• Transliterate Endpoint:

  POST /convert
  Content-Type: application/json
  
  {
    "text": "tmi kmn aso? amr khub vlo lagse."
  }
  

  Response:

  {
    "original": "tmi kmn aso? amr khub vlo lagse.",
    "converted": "তুমি কেমন আছো? আমার খুব ভালো লাগছে."
  }
  

• Stats Endpoint:

  GET /stats
  

  Returns dictionary size, active rules, and baseline mappings count.

📄 License

Distributed under the MIT License. See LICENSE for details.