home · mobile · calendar · colloquia · 2007-2008 · 

Colloquium - Riloff

Desiderata for Annotating Data to Train and Evaluate Bootstrapping Algorithms
University of Utah

In this talk I will address a variety of issues surrounding the potential benefits, challenges, and possible pitfalls of annotating data for bootstrapping algorithms. The discussion will focus both on issues associated with the creation of annotated training data as well as issues associated with the creation of annotated evaluation data. In supervised learning scenarios the same types of manual annotations are typically used both for training and evaluation, but this is not necessarily the case for bootstrapping algorithms. I will argue that it may be desirable to adopt different strategies for these cases, and that annotating data for evaluation purposes should be a major focus of manual annotation efforts.

Sponsored by the Computational Language and Education Research Center (CLEAR).
Hosted by Martha Palmer.

Department of Computer Science
University of Colorado Boulder
Boulder, CO 80309-0430 USA
May 5, 2012 (14:13)