Facebook has released a dataset designed to test multilingual translation models. The dataset, known as Flores-101, contains 3,001 English sentences taken from Wikipedia, and their translated counterparts, in 101 human languages. Researchers can use the data to evaluate translation systems’ performance and advance natural language processing projects.
Image credit: Flickr user Mike
