Detecting Social Media Manipulation in Low-Resource Languages