Logo Lanfrica

HamzaBouhassou/multilingual-sentiment-analysis

Domaine:

natural language processing

Type de record:

project
Créateur:
Ham
Hôte:
Sentiment analysis of YouTube comments in Arabic, Darija, French & English using NLP and ML # Multilingual Sentiment Analysis — YouTube Comments Sentiment analysis system for YouTube comments in **Arabic, Darija, French & English**, applied to Hespress trending videos. ## Overview - Scraped 2,500+ comments via YouTube Data API v3 - Multi-language detection: Arabic · Darija · French · English - MongoDB storage pipeline - 6 ML classifiers compared: Logistic Regression, SVM, LDA, QDA, KNN, Decision Trees - **Best accuracy: Decision Tree — 74.6%** - Sentiment distribution: 63.1% Positive · 36.9% Negative ## Pipeline Data Collection → Language Detection → Preprocessing → Labeling → Model Training → Evaluation ## Results ## Tech Stack - Python · Scikit-learn · NLTK · MongoDB · YouTube Data API v3 - langdetect · TF-IDF Vectorization ## Files | File | Role | |------|------| | `scrap.py` | YouTube API scraping + language detection | | `pre-process.py` | Comment cleaning and preprocessing | | `filtre_fr_ang.py` | French/English filtering + labeling | | `filtredar_ab_.py` | Arabic/Darija filtering + labeling | | `class.py` | ML classification (6 models) | | `stat_*.py` | Statistics and visualizations | ## Author **Hamza Bouhassou** — Master AI & Data Science @ Istanbul Aydın University