•   When: Friday, February 19, 2021 from 03:00 PM to 04:00 PM
•   Speakers: Antonios Anastasopoulos
•   Export to iCal

Active learning (AL) uses a data selection algorithm to select useful training samples to minimize annotation cost. This is now an essential tool for building syntactic analyzers such as part-of-speech (POS) taggers in low-resource languages.

In this talk, I will discuss how commonly used data selection heuristics are far from optimal, and how we can pose the problem of AL as selecting instances which maximally reduce the confusion between outputs, a strategy that significantly outperforms other AL strategies. I will also go through our efforts in building linguist-in-the-loop tools for language documentation.