constit parser desc - apache/ctakes GitHub Wiki

The model in parsermodel is trained on Mipacq data (clinical notes, clinical questions, medpedia, etc.) as well as the Penn Treebank (Wall Street Journal). This model performs about 1.5 better labeled F-score on some preliminary runs. fastModel is trained only on clinical data but is a much smaller model and runs a lot faster.