Christian Federmann

About

I focus on large-scale human evaluation for machine translation, working with the Microsoft Translator team at Microsoft AI+Research.

On March 14, 2018, we released our research paper “Achieving Human Parity on Automatic Chinese to English News Translation (opens in new tab)“. My contributions include a definition of human parity for translation, and results from a large-scale human evaluation campaign which empirically and statistically proved that our research systems are indistinguishable from human translations on various subsets of the newstest2017 test set from WMT17.

We have released translation output, new human references, and all data collected in our human evaluation campaigns here: aka.ms/Translator-HumanParityData (opens in new tab).