TurboParser 2.1 was used as a base for all experiments. On top, the following patches were applied:

Due to the difference in the evaluation system (namely the evaluation of incremental output), the I/O interface of TurboParser was changed to read and write one file per sentence.

The incremental gold standard for both German and English was created using the create_prefixes_virtual_unified.py script from our toolbox.
  • The language-dependent parts are located in transform_![de|en].py
  • The evaluation script is cdgevaluator_timecourse.py
  • * incr-tp-shell-scripts.tar.gz: Files that describe (more or less) how each experiment was carried out are here:

-- ArneKoehn - 18 Sep 2014

This site is powered by FoswikiCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding Foswiki? Send feedback