Can We Catch Bad Translations Automatically?

Machine translation (MT) is great when it works, but it fails often and in unpredictable ways. What causes bad machine translations and can we catch the resulting issues?  What about bad human translations? Providers and enterprises like Unbabel, Microsoft, Amazon and eBay use "quality estimation", the automatic prediction of translation quality and risk, to: - Estimate post-editing effort - Safely auto-approve raw machine translation for as many segments as possible - Focus on LQA In this GALA webinar Adam Bittlingmayer discusses how this technology works and how reliable it is. He also covers where can you get this technology and how can you integrate it into your workflow today. For more information about GALA: https://www.gala-global.org/ For more on globalization and localization news, subscribe to our newsletter: https://bit.ly/3I2FRmq For more on GALA events: https://bit.ly/3FDKZNY Chapters 02:48 Introduction by the speaker 03:29 Why catch bad translations? 07:12 Can we catch bad translations? 09:50 Hybrid translation 14:13 Demo 29:49 Who has this technology and how to get it? 39:26 Q&A Speaker Adam Bittlingmayer is CEO and co-founder at ModelFront. ModelFront provides translation risk prediction. He is a language learner and software engineer with experience at Google Translate as well as Adobe, Google Play and startups. Beyond translation, his interests include input correction, language identification, transliteration and synthetic data generation. #localization #machinetranslation #languagequality