Bootstrapping and Linguistics
Tuesday, February 25, 2003, 01:00pm - 02:00pm
University of Michigan, Computational Linguistics
An area of current interest in computational linguistics is semi-supervised learning, or bootstrapping. Two of the most popular bootstrapping algorithms are co-training and the Yarowsky algorithm. I discuss why these algorithms work. I also discuss what they have to do with linguistics, suggesting in particular how bootstrapping methods can be used to aid the analysis of undocumented languages.