Although the prediction of dialects is an important language processing task, with a wide range of applications, existing work is largely limited to coarse-grained varieties. Inspired by geolocation research, we propose the novel task of Micro-Dialect Identification (MDI) and introduce MARBERT, a new language model with striking abilities to predict a fine-grained variety (as small as that of a city) given a single, short message. For modeling, we offer a range of novel spatially and linguistically-motivated multi-task learning models. To showcase the utility of our models, we introduce a new, large-scale dataset of Arabic micro-varieties (low-resource) suited to our tasks. MARBERT predicts micro-dialects with 9.9% F1, ~76X better than a majority class baseline. Our new language model also establishes new state-of-the-art on several external tasks.

TRIPOLI – A drone airstrike by eastern Libyan forces on the southern Libyan town of Murzuq has killed at least 43 people, a local official said on Monday. The attack is the second major airstrike blamed on the eastern Libyan National Army (LNA) forces loyal to Khalifa Hifter after at least 44 migrants were killed in June when a detention center in a suburb of the capital Tripoli was hit. The LNA confirmed a strike late on Sunday on Murzuq, but denied it had targeted any civilians. The LNA had also denied it had hit the detention center but acknowledged increased air strikes on the capital. The internationally recognized government based in Tripoli opposing Hifter said dozens were killed and wounded in Murzuq. Reached by telephone, Murzuq municipal council member Mohamed Omar told Reuters: "The airstrike resulted in 43 killed and 51 wounded.