⥠Multi-Script Alignment Engine
Type or paste any formal Bengali or English sentence to produce Natural Avro Banglish and Urban Code-Mixed text simultaneously.
Quick Presets:
Apple & DNA Sequencing
Apollo 11 & Armstrong
Two-Factor Auth (2FA)
Bleaching Powder Warning
English Query (All-in-One)
Natural Avro Banglish
Apple Inc-er CEO Tim Cook-er shompurno DNA sequencing prodan korun.
No robotic vowel omissions (keno, na, lojja preserved).
Modern Urban Code-Mixed
Apple Inc-āĻāĻ° CEO Tim Cook-āĻāĻ° āĻ¸āĻŽā§āĻĒā§āĻ°ā§āĻŖ DNA āĻ¸āĻŋāĻā§āĻ¯āĻŧā§āĻ¨ā§āĻ¸āĻŋāĻ āĻĒā§āĻ°āĻĻāĻžāĻ¨ āĻāĻ°ā§āĻ¨āĨ¤
Preserves Bengali morpho-syntax while keeping authentic technical loanwords.
đ Upload & Custom Multi-Script Dataset Exporter
Upload your dataset in CSV, JSON, or JSONL format. Supports English prompts, Bengali prompts, or Mixed prompts. Select columns and choose your desired output formats.
Click to Browse or Drag & Drop Dataset File
Supports .csv, .json (Array of Objects), or .jsonl files (English or Bengali)
| Status |
|---|
| Upload a CSV or JSON file and click 'Process & Generate Output' to see preview. |
đ Script & Language Identification (LID)
Classify whether social media text, comments, or scraped corpora are Bengali, Banglish, English, or Code-Mixed before training LLMs.
Bengali Sample
Banglish Sample
Code-Mixed Sample
English Sample
Predicted Category
Avro Banglish
Detected high-frequency Bengali phonetic tokens written in Latin script.
Unicode Distribution Breakdown
Bengali: 0.0% | Latin: 86.4% | Digits: 0.0% | Punctuation/Spaces: 13.6%
𧚠Social Media Noise & Text Normalizer
Removes Zero-Width Non-Joiner (ZWNJ) artifacts, normalizes punctuation, and collapses social media letter elongation and SMS slang.
ZWNJ & Duplicate Dari
Banglish Elongation & Slang
MT Acronym Trash Loop
Cleaned Text
onekta bhalo lagche keno?
đ¤ 1-Click LLM Alignment Single Exporter
Quick single-pair generator for Unsloth, LLaMA-Factory, Axolotl, or Hugging Face TRL.