How brightprogrammer7/deepfakenews works
Project type and purpose
• This is a Python-based AI pipeline/service (no frontend), “AIHakikaNews,” whose goal is to scrape news data, process it via NLP, assess sentiment (“Feeling”), and detect fake news using a trained model.
• Key indicators:
• app.py entry point
• “Scraping” directory with raw data
• “NLP” for vectorization models
• “Feeling” for sentiment
• “FakeNews” for the classification model
File structure
• Top-level files:
• .gitignore, README.md, app.py, venv
• Major directories (each with its own __pycache__):
• Scraping: scraping.py + raw data files
• NLP: nlp.py + prebuilt models (bagOfWords, tfidf, word2vec)
• Feeling: feeling.py (sentiment analysis)
• FakeNews: fakeNews.py + SVC.model (fake news classifier)
• No frontend folder ⇒ this is a back-end processing service.
Extract insights from README
a. Main components
1. Data Ingestion (Scraping)
2. Data Storage (raw .txt/.csv files under Scraping/data & FakeNews/data.csv)
3. Preprocessing & Feature Extraction (NLP module)
4. Sentiment Analysis (Feeling module)
5. Fake News Detection (FakeNews module)
6. Orchestration/API Layer (app.py)
7. Environment (Python venv)
b. Relationships & interactions
• Scraper → raw data files
• Raw data → NLP preprocessing
• NLP outputs → two parallel branches:
– Feeling → sentiment score
– FakeNews → classification label
• app.py calls each module in sequence and aggregates results
c. Architectural patterns & principles
• Modular pipeline: each stage in its own directory
• Separation of concerns: scraping, preprocessing, sentiment, classification
• Reusable models: pretrained assets under NLP/models and FakeNews/SVC.model
d. Key technologies
• Python 3
• Libraries (inferred): scikit-learn (SVC, tfidf), gensim or similar for word2vec
• Local file I/O for data exchange