The Fantastic Best-quality Engines and How to Find Them: Evaluating MT in 2020
Track: TAUS | TS4 | Advanced |
Friday, November 8, 2019, 1:30am – 2:15am
Held in: California
Silvio Picinini - eBay
Konstantin Savenkov - Intento, Inc.
Kirill Soloviev - ContentQuo
Andrea Tabacchi - Memsource
Host: Anne-Maj van der Meer
With neural machine translation evolving at a maddening tempo, it’s difficult even for larger companies to keep track of it. What methodologies and processes should we use in 2020 to assess and predict the quality of machine translation (MT) engines at scale across languages, content types and verticals? How do we combine automatic metrics, human evaluation, quality estimation methods and end-user input to give us the right granularity and reliability of data on our MT quality? How do we ensure all these processes are efficient and lead to solid, data-informed decisions? Join our panel discussion to find out.
Takeaways: Attendees will learn the current state-of-the-art of various MT quality evaluation methods; understand the strengths, weaknesses, costs and benefits of each method; and figure out how to make better decisions when selecting MT engines for production deployment.