Passport Translator

A Flask-based web application for translating Ukrainian international passports to Turkish, multilingual translation documents and images with additional features like named entity recognition (NER).

Features

1. Ukrainian Passport Translation: Translates Ukrainian international passports into Turkish.
2. General Document Translation: Translates documents and images from any language into English, with named entity recognition using SpaCy.
3. Document Scanning: Detects and extracts document boundaries from uploaded images.
4. Changelog: Tracks and displays changes made to translations.
5. Email Notifications: Sends emails with custom content.

Technologies Used

• Backend: Python, Flask
• Frontend: HTML, CSS, SCSS, JS
• Libraries:
  • OpenCV for image processing
  • SpaCy for natural language processing
  • Mistral AI for image-to-text conversion
  • Flask for web framework
  • ReportLab for PDF generation
  • NumPy for numerical operations
  • smtplib for email handling

Installation

1. Clone the repository:

git clone https://github.com/your-username/passport_translator.git
cd passport_translator

2. Create a virtual environment and activate it:

python -m venv venv
source venv/bin/activate  # On Windows use `venv\Scripts\activate`

3. Install the required packages:

1. Requirements.txt:

pip install -r requirements.txt

2. Download the SpaCy model:

python -m spacy download en_core_web_trf

3. Install PyTorch with GPU support (if applicable): Visit the PyTorch installation page and follow the instructions to install the appropriate version for your system. For example:

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

4. Set up environment variables in a .env file:

EMAIL_BACKEND=smtp
EMAIL_HOST=smtp.example.com
EMAIL_PORT=587
EMAIL_HOST_USER=your-email@example.com
EMAIL_HOST_PASSWORD=your-password
EMAIL_USE_TLS=True

5. Run the application:

python main.py

Usage

1. Ukrainian Passport Translation:

• Upload an image of a Ukrainian passport.
• Choose the points that form a document boundaries to be translated.
• The application will translate the text into Turkish and generate you a pdf file.

2. General Document Translation:

• Upload an image or document in any language.
• Choose the points that form a document boundaries to be translated.
• The application will translate the text into English and highlight named entities.

Changelog Feature

There's a changelog feature that tracks changes made to translations for documents and pictures. You can view the changelog in the application.

Email Notifications

The application can send email notifications with custom content. Make sure to configure your email settings in the .env file.

Named Entity Recognition (NER) with SpaCy

The application uses the spaCy English trf model (en_core_web_trf) for named entity recognition. Below are the entity labels recognized by the model:

Model Performance Metrics (en_core_web_trf)

Metric	Value	Description
ENTS_P	90.08%	Precision: How many of the detected entities were correct
ENTS_R	90.30%	Recall: How many of the actual entities were correctly detected
ENTS_F	90.19%	F1-score: Harmonic mean of precision and recall

Label	Description	Examples
CARDINAL	Numbers without units	"three", "1000", "millions"
DATE	Calendar dates	"January 1", "2024", "last year"
EVENT	Historical, cultural, or events	"World War II", "the Assumption of the Blessed Virgin Mary"
FAC	Infrastructure objects	"Golden Gate Bridge", "The Bakhchisaray Palace"
GPE	Countries, cities, regions	"Ukraine", "Crimea", "California"
LANGUAGE	Names of languages	"English", "Ukrainian"
LAW	Legal documents, statutes	"Constitution", "Article 5", "The Civil Rights Act"
LOC	Natural locations	"Mount Everest", "the Crimean Mountains", "Sahara"
MONEY	Amounts of money with currency	"$50", "10 euros", "one million yen"
NORP	Nationalities, religious groups	"Ukrainians", "Christians", "Slavic"
ORDINAL	Positions in a sequence	"first", "2nd", "third"
ORG	Companies, institutions	"Google", "UN", "Harvard University"
PERCENT	Percent values	"5%", "thirty percent"
PERSON	Given names, surnames	"George Washington", "Bruno Pelletier"
PRODUCT	Manufactured objects	"iPhone", "Tesla Model S", "PlayStation"
QUANTITY	Numbers with measurement units	"5 kilograms", "30 miles"
TIME	Specific times of day	"2 PM", "midnight", "noon"
WORK_OF_ART	Titles of books, films, paintings	"Mona Lisa", "1000 and One Nights", "War and Peace"