To Translate or Not to Translate — A Hybrid Automation Approach Using AI to Enrich Translation Context

Track: Multilingual AI | TA5 |
Wednesday, June 9, 2021, 3:45pm – 4:30pm
Held in: Jujama 2
Agustín Da Fieno Delucchi - Microsoft 
Rafa Moral - Lionbridge

In this session, we will present a case study of a hybrid methodology for detecting and classifying known entities or common issues on text segments, including do-not-translate (DNT) segments, named entities, typos, and so on. This hybrid model involves rule-based, statistical, and machine learning methodologies that, combined, provide apparent efficiencies to the translation pipeline.

Takeaways: Attendees will have the opportunity to see a real-world application of AI solutions in localization, in combination with commonly used rule-based and statistical methodologies, and in the specific context of the translation process.