ammaromari/arabic-nlp-research

Cross-lingual transfer learning for Arabic dialects. Addresses low-resource NLP challenges in MENA region with transformer based approaches and dialect specific benchmarks. # Arabic NLP Research Cross-lingual transfer learning for Arabic dialects. Addresses low-resource NLP challenges in the MENA region using character-level n-gram features and transformer-based approaches for dialect identification and classification. ## Dialects Supported - Modern Standard Arabic (MSA) - Egyptian Arabic - Gulf Arabic (Saudi Arabia, UAE, Kuwait) - Levantine Arabic (Syria, Lebanon, Palestine) - Maghrebi Arabic (Morocco, Algeria, Tunisia) ## Features - Character-level n-gram TF-IDF feature extraction - Multi-class dialect classification with confidence scoring - Logistic Regression and Random Forest baselines - 5-fold cross-validation evaluation - Live inference with probability visualization - Preprocessing pipeline for Arabic text normalization ## Quick Start git clone github.com cd arabic-nlp-research python3 -m venv venv source venv/bin/activate pip install -r requirements.txt python3 src/models/arabic_dialect_classifier.py ## Research Context Arabic dialect identification is a critical component of MENA-focused NLP systems. With 400M+ Arabic speakers across 22 countries using mutually unintelligible dialects, robust dialect identification enables more accurate downstream NLP tasks including sentiment analysis, machine translation, and content moderation. ## License MIT