OPINAR: A Multi-Source Arabic Opinion Dataset with Keywords, 2004–2025

OPINAR is a large-scale Arabic opinion-article dataset covering 2004–2025, containing 280,890 articles, approximately X million tokens, and Y million unique word types, collected from 85 news websites across 21 countries (15 Arab and 6 non-Arab). Sources were selected primarily for providing keyword metadata. Each article includes publication date, source, ownership country, keywords/topics when available, and source URL.

OPINAR is a large-scale Arabic opinion-article dataset covering 2004–2025, containing 280,885 articles, approximately 181 million tokens, and 3.1 million unique word types, collected from 84 news websites across 22 countries (16 Arab and 6 non-Arab). Sources were selected primarily for providing keyword metadata. Each article includes publication date, source, ownership country, keywords/topics when available, and source URL.

The dataset is organized hierarchically by country → source → year → month, and each article is assigned a unique global identifier (OPN_XXXXXXXX). OPINAR supports research in opinion mining, sentiment analysis, discourse analysis, media studies, and Arabic NLP, offering extensive multi-source opinion content with transparent metadata.