How much morphology do we need to parse German?
Time: 01:00pm - 02:00pm
Place: Ballantine Hall 141
Daniel Dakota, Dan Whyatt and Sandra KueblerPOS tagging and parsing are generally considered in a pipeline solution where the POS tagger provides the input for the parser. However, this raises the question how much information the POS tagger should provide: If we have a small tagset, the POS tagging task is easy and can be performed reliably, but the parser may have too little information to make accurate decisions. If we include morphological information, the parser found profit from that, but then the information from the POS tagger will be less reliable. We will investigate which POS tag level is the most accurate, in terms of POS tagging accuracy as well as parsing.
In category: Computational linguistics
Natural Language Toolkit
Time: 01:00pm - 04:00pm
Place: Wells Library Information Commons Cluster #3
JEvents v3.0.9 Stable
Copyright © 2006-2013
Powered by Joomla!®
Copyright © 2008 The Trustees of Indiana University | Copyright Complaints