Multilingual MENA News Events Data | Arabic, Turkish, Farsi, English | 24 Countries, 100+ Sources

The provider’s MENA Geopolitical News Dataset is a structured, labelled corpus of 55,000+ news items aggregated from 100+ sources across 24 countries in the Middle East and North Africa region. Unlike raw web-scraped text, every item in this dataset has been processed through proprietary fine-tuned machine learning models to produce consistent, reliable annotations, making it immediately usable for LLM training, fine-tuning, and evaluation without additional labelling overhead. What's included: Each item contains a unique ID, title, URL, source domain, source country, source language, published timestamp, AI-generated summary, topic category, category confidence score, reliability score, reliability score breakdown, relevant countries, and official source flag; 19 structured fields per item in consistent JSON format across all four languages. Topic classification: Every item is classified into one of six categories (conflict, politics, economy, energy, innovation, or culture) using a fine-tuned XLM-RoBERTa-base model achieving 95.8% macro F1 accuracy across all six categories on held-out test data. The model was trained specifically on MENA geopolitical news content and handles Arabic, Turkish, Farsi, and English natively without translation. Source reliability scoring: Each item carries a reliability score from 0–100 derived from a fine-tuned DeBERTa-v3-base model that detects seven independent signals: official sourcing (automatic score of 100), regional specialist sourcing, spec