erickrf/nlpnet

training the dependency parser

DeepInEvil opened this issue · 1 comments

Hi,

How can we train the dependency parser only?
also the data you have used to train the dependency parser, is that an open datasource ?

Regards,
Debanjan

Hi,

I haven't given much attention to documenting the dependency parser functionality since its results were below expected, but you can get the command line instructions running nlpnet-train.py dependency -h.

It has two stages: the unlabeled (finding the dependency edges) and labeled (labeling the previously found edges), and they are trained separately. Instructions for each are available with nlpnet-train.py dependency unlabeled -h and nlpnet-train.py dependency labeled -h