CoNLL-2004 and CoNLL-2005 Shared Tasks

Semantic Role Labeling: F.A.Q.

This is a collection of FAQs that has been compiled after the celebration of the Shared Task in CoNLL-2005. They address questions and issues that are relevant for using the CoNLL experimental setting for Semantic Role Labeling.

• I want to do experiments with my new Semantic Role Labeling system. Can I use the data and software that is available here?
  
  Yes. We strongly encourage you to follow the setting of the CoNLL-2005 shared task. It is much more complete than that of the shared task in 2004. Importantly, it has been used by many researchers after the conference, so you will be able to compare your results with those of other teams. Please, check out the section on Data and Software.
  
  
  
• What sections of the Brown corpus correspond to the "Test Brown" set in the Shared Task?
  
  The "Test Brown" of the CoNLL-2005 evaluation corresponds to the sentences in the files ck01.mrg, ck02.mrg and ck03.mrg of the Brown corpus. These files are included in the Penn Treebank release 3. The "ck" files in Brown account for "general fiction" texts.