Nationality identication unlocks important demographic information, with many applications in biomedical and sociological research. Existing name-based nationality classiers use name substrings as features and are trained on small, unrepresentative sets of labeled names, typically extracted from Wikipedia. As a result, these methods achieve limited performance and cannot support ne-grained classication. We exploit the phenomena of homophily in communication patterns to learn name embeddings, a new representation that encodes gender, ethnicity, and nationality which is readily applicable to building classiers and other systems. rough our analysis of 57M contact lists from a major Internet company, we are able to design a ne-grained nationality classier covering 39 groups representing over 90% of the world population. In an evaluation against other published systems over 13 common classes, our F1 score (0.795) is substantial beer than our closest competitor Ethnea (0.580). To the best of our knowledge, this is the most accurate, ne-grained nationality classier available. As a social media application, we apply our classiers to the followers of major Twier celebrities over six dierent domains. We demonstrate stark dierences in the ethnicities of the followers of Trump and Obama, and in the sports and entertainments favored by dierent groups. Finally, we identify an anomalous political gure whose presumably inated following appears largely incapable of reading the language he posts in
mobile