Built a Text to Text Translator using NLTK and Transformers.
- Supports Translation of English to Bengali, Tamil, Telugu, Gujarati, Marathi and Hindi.
- Uses BanglaT5 which achieved an exceptional score of 25.2 on SacreBLEU metric while mt5 (Industry Standard) scored much lower at 22.5
Future Work:
- Adding functionality of uploading Images and Files
- OCR will run on these files and provide translation automatically
- Run
python -m venv .venv
- Run
.venv/Scripts/Activate
- Run
pip install -r requirements.txt
- Get your
HuggingFace Token
fromhttps://huggingface.co/settings/tokens
- Generate a
Write
Token and store this in a file namedsecret.py
- Run
python app.py