[Corpora-List] POS-tagger maintenance and improvement

Oliver Mason O.Mason at bham.ac.uk
Wed Feb 25 21:20:03 CET 2009


I have a while ago started to add a rule-based pre/post-processor to QTag, which is meant to iron out some typical systematic errors that often occur with statistical tagging. While this does improve accuracy, it makes the tagger more dependent on a particular tagset (and also a particular language). However, it might still be useful to have those rules in a language/tagset specific resource file rather than applicable to the tagger itself.

So in principle there should be no problem in adding more rules to the tagger to cater for systematic errors without having to mess about with the training data, it's only a question of time and effort.

Best wishes, Oliver

More information about the Corpora mailing list