Data for testing ranking code.

The files in this directory contain extracted features for doing parse
selection on the Redwoods treebank. There are four files, each
corresponding to a different section of the treebank, and each
containing features for sentences with 10 parses or less. The feature
set is the "DT-NGRAM" feature set on derivation trees, discussed in
the following paper:


  Jason Baldridge and Miles Osborne. Active Learning and Logarithmic
  Opinion Pools for HPSG Parse Selection. Natural Language Engineering
  13(1):1-32, Cambridge, UK.
  
  http://comp.ling.utexas.edu/jbaldrid/papers/BaldridgeOsborneNLE.pdf

The vm6 and vm32 sections are from similar domains (travel planning
and appointment scheduling), and the ecoc and ecpa are both from
ecommerce domains.