* Don't consider NPs connected to parse via conj relation as noun chunks. Change motivated by the nested noun chunks identified in Issue #203, but might be problematic. Also allow root NPs to be considered noun chunks.

This commit is contained in:
Matthew Honnibal 2016-01-16 17:52:40 +01:00
parent 4a16dbfeca
commit fc8f26584a

View File

@ -241,7 +241,7 @@ cdef class Doc:
"to install the data")
cdef const TokenC* word
labels = ['nsubj', 'dobj', 'nsubjpass', 'pcomp', 'pobj', 'attr', 'conj']
labels = ['nsubj', 'dobj', 'nsubjpass', 'pcomp', 'pobj', 'attr', 'root']
np_deps = [self.vocab.strings[label] for label in labels]
np_label = self.vocab.strings['NP']
for i in range(self.length):