Rooba
/
spaCy
mirror of https://github.com/explosion/spaCy.git
1
1
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Adjust text classification model

This commit is contained in:
Matthew Honnibal 2017-09-02 11:41:00 +02:00
parent ac040b99bb
commit a824cf8f9a
1 changed files with 15 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

23
spacy/_ml.py
View File

@ -24,7 +24,7 @@ from thinc.linear.linear import LinearModel
from thinc.api import uniqued, wrap, flatten_add_lengths from thinc.api import uniqued, wrap, flatten_add_lengths
from .attrs import ID, ORTH, LOWER, NORM, PREFIX, SUFFIX, SHAPE, TAG, DEP from .attrs import ID, ORTH, LOWER, NORM, PREFIX, SUFFIX, SHAPE, TAG, DEP, CLUSTER
from .tokens.doc import Doc from .tokens.doc import Doc
from . import util from . import util
@ -521,25 +521,30 @@ def build_text_classifier(nr_class, width=64, **cfg):
trained_vectors = ( trained_vectors = (
FeatureExtracter([ORTH, LOWER, PREFIX, SUFFIX, SHAPE, ID]) FeatureExtracter([ORTH, LOWER, PREFIX, SUFFIX, SHAPE, ID])
>> with_flatten( >> with_flatten(
uniqued(
(lower | prefix | suffix | shape) (lower | prefix | suffix | shape)
>> LN(Maxout(width, 64+32+32+32)),
column=0
)
) )
) )
convolution = ( static_vectors = (
ExtractWindow(nW=1) SpacyVectors
>> LN(Maxout(width, width*3)) >> with_flatten(Affine(width, 300))
) )
cnn_model = ( cnn_model = (
# TODO Make concatenate support lists # TODO Make concatenate support lists
concatenate_lists(trained_vectors, SpacyVectors) concatenate_lists(trained_vectors, static_vectors)
>> with_flatten(
LN(Maxout(width, 64+32+32+32+300))
>> convolution ** 4, pad=4)
>> flatten_add_lengths >> flatten_add_lengths
>> with_getitem(0,
SELU(width, width*2)
>> (ExtractWindow(nW=1) >> SELU(width, width*3)) ** 2
)
>> ParametricAttention(width) >> ParametricAttention(width)
>> Pooling(sum_pool) >> Pooling(sum_pool)
>> ReLu(width, width) >> SELU(width, width) ** 2
>> zero_init(Affine(nr_class, width, drop_factor=0.0)) >> zero_init(Affine(nr_class, width, drop_factor=0.0))
) )