Take Shelter, Zanmi: Digitally Alerting Cyclone Victims in Their Languages

Natural disasters such as tropical cyclones cause annual devastation and take a heavy social cost, as disadvantaged communities are typically hit hardest.Among these communities are the speakers of minority and low-resource languages, who may not be sufficiently informed about incoming weather events to prepare.This work presents an analysis of the current state of machine translation for natural disasters in the languages of communities that are threatened by them.Results suggest that commercial systems are promising, and that in-genre fine-tuning data are beneficial.