Neural AutoReview — Using Neural Networks for Automated Review and Grammatical Error Correction

Track: Sponsored Session | SS10A |
Thursday, June 10, 2021, 11:15am – 12:00pm
Held in: Jujama 1
Presenter:
Joern Wuebker - Lilt
SPONSORED BY: Lilt

For several years, interactive adaptive neural machine translation has been successfully applied at Lilt to augment professional linguists in a human-in-the-loop setting, allowing them to work faster and produce higher quality output. In this session, we will discuss using similar AI technology to build customized quality assurance (QA) models that learn from reviewers’ corrections to the first pass translation and can provide high-precision corrections that traditional QA models are unable to capture.

Takeaways: Attendees will see the neural AutoReview prototype in action and get an overview of the types of reviewer corrections that can be automated. They will also be given additional background on how the technology works, and learn about the evaluation methodology we use to analyze its effectiveness.