Arab Computational Propaganda on Twitter

The database includes five datasets. Three datasets were extracted from a dataset published by Twitter that includes tweets from malicious accounts trying to manipulate public opinion in the Kingdom of Saudi Arabia. We focused on sports and banking topics when extracting data. Although the propagandist tweets were published by malicious accounts, as Twitter stated, the tweets at their level were not classified as propaganda or not. Propagandists usually mix propaganda and non-propaganda tweets in an attempt to hide their identities. Therefore, it was necessary to classify their tweets as propaganda or not, based on the propaganda technique used. Since the datasets are very large, we annotated a sample of 2,100 tweets. As for reliable account data, we were keen to identify reliable Saudi sources. Then, their tweets that discussed the same topics discussed by the malicious users were crawled. There are two datasets for reliable users in sports and banking topics. The dataset is made up of 16,355,558 tweets from propagandist users and 156,524 tweets from reliable users for the time period of January 1, 2019, to December 31, 20202.