These files contain the train and test data for for the three parts of 
the CoNLL-2001 shared task:

   testa1: development data part 1
   testa2: development data part 2
   testa2: development data part 3
   testb1: test data part 1
   testb2: test data part 2
   testb3: test data part 3 (corrected version: 20030803)
   testb3.org: test data part 3 (original version used in pre-2003 publications)
   train1: train data part 1
   train2: train data part 2
   train3: train data part 3

Relevant publication:

   Erik F. Tjong Kim Sang and Hervi Dijean, Introduction to the CoNLL-2001
   Shared Task: Clause Identification. In: Proceedings of CoNLL-2001,
   Toulouse, France, 2001.

Associated url: http://lcg-www.uia.ac.be/conll2001/clauses/

Note: the original testb3 files contained duplicate clauses: (S(S words S)S).
This data set has been replaced by a corrected version on August 3, 2003.
All publications before 2003 refer to the original data sets and report
recall and F1 measures which are lower than they should be.