Home BlogDataset Training Multilingual Language Models

Training Multilingual Language Models

by Morgan Stevens
by
Speech bubbles

Amazon has created a dataset of common phrases in 51 languages to train massive multilingual natural language understanding models. The dataset contains 19,521 common phrases such as “what is the weather in New York City?” in 51 languages. Researchers can use the dataset to train a single machine learning model to understand the same phrase in all 51 languages. 

Get the data.

Image credit: Flickr user Marc Wathieu

You may also like

Show Buttons
Hide Buttons