Home BlogDataset Expanding Translation Systems’ Capabilities

Expanding Translation Systems’ Capabilities

by Morgan Stevens
A spherical art installation displaying the word welcome in several languages

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.

Get the data.

Image credit: Flickr user Mike

You may also like

Show Buttons
Hide Buttons