Fact_Check (PAK CHECKER)

A small fact-checking app that can:

• Train a Naive Bayes model using a bag-of-words vocabulary
• Evaluate the model
• Classify new news from:
  • manual text (title + body)
  • a URL (scraping)
  • an image (OCR)
• Batch classify from a CSV
• Scrape multiple URLs and store results

Requirements

• Python 3.9+

Install dependencies:

pip install -r requirements.txt

Note: The OCR part uses Tesseract (pytesseract). Ensure Tesseract is installed on your system.

Project structure (high level)

• run_cli.py / run_gui.py: entry points
• src/cli.py: CLI menu
• src/gui.py: Tkinter GUI
• src/fake_news_detector.py: chooses CLI vs GUI based on --cli/--gui
• src/training.py: training logic
• src/evaluation.py: evaluation logic
• src/verifier.py: verifies/classifies an input (optionally across multiple seeds)
• src/scraper.py: scraping helpers
• src/ocr_reader.py: OCR extraction

How to run

Option A: CLI

From the repo root:

python run_cli.py

You’ll see a menu with options like:

• Train model
• Evaluate model
• Classify new article (manual / URL / image)
• Batch classify from CSV
• Scrape news (multiple URLs)
• Export Bag-of-Words CSV for last seed
• Reload stopwords

Option B: GUI

python run_gui.py

Tabs:

• Detect News: classify text / URL / image (OCR)
• Scraping: paste URLs and scrape into data/true.csv or data/false.csv
• Training & Settings: train/retrain, evaluate, export vocabulary, reload stopwords

Option C: Single entry via arguments

python -m src.fake_news_detector --cli
python -m src.fake_news_detector --gui

(If your Python import path differs, use run_cli.py / run_gui.py instead.)

Classify examples

1) Manual text classification (CLI)

In the CLI menu:

1. Choose 3) Classify new article
2. Select mode 1) manual
3. Enter title and text

The app will only mark results CONFIDENT when models agree across configured seeds.

2) URL classification (CLI)

In the CLI menu:

• Choose 3) Classify new article
• Select mode 2) URL

3) Image classification (OCR)

• Choose 3) Classify new article
• Select mode 3) image
• Select an image file (.png/.jpg/.jpeg/.bmp)

Data files

The app uses the following CSVs:

• data/true.csv
• data/false.csv
• data/stopwords.csv

It also writes logs into:

• logs/user_inputs.csv
• logs/scraped_log.csv
• logs/metrics_history.csv
• logs/training_log.csv

Notes / troubleshooting

• If OCR fails:
  • confirm Tesseract is installed
  • make sure pytesseract can find the Tesseract executable
• If you see missing-model/vocabulary errors:
  • run Train model first (CLI menu option 1 or GUI training tab)

License

Add your license here.