explosion/spaCy
1
1
Fork 0
mirror of https://github.com/explosion/spaCy.git synced 2025-06-01 11:43:03 +03:00
Code Issues Packages Projects Releases Wiki Activity

Bug fixes to pipeline

Browse Source
This commit is contained in:
Matthew Honnibal 2017-05-18 04:29:51 -05:00
parent 8815507f8e
commit b460533827
1 changed files with 25 additions and 22 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

45
spacy/pipeline.pyx
View File

@ -8,6 +8,7 @@ from thinc.neural import Model, Softmax
import numpy import numpy
cimport numpy as np cimport numpy as np
import cytoolz import cytoolz
import util
from thinc.api import add, layerize, chain, clone, concatenate from thinc.api import add, layerize, chain, clone, concatenate
from thinc.neural import Model, Maxout, Softmax, Affine from thinc.neural import Model, Maxout, Softmax, Affine
@ -42,12 +43,14 @@ class TokenVectorEncoder(object):
@classmethod @classmethod
def Model(cls, width=128, embed_size=5000, **cfg): def Model(cls, width=128, embed_size=5000, **cfg):
width = util.env_opt('token_vector_width', width)
embed_size = util.env_opt('embed_size', embed_size)
return Tok2Vec(width, embed_size, preprocess=None) return Tok2Vec(width, embed_size, preprocess=None)
def __init__(self, vocab, model=True, **cfg): def __init__(self, vocab, model=True, **cfg):
self.vocab = vocab self.vocab = vocab
self.doc2feats = doc2feats() self.doc2feats = doc2feats()
self.model = self.Model() if model is True else model self.model = model
def __call__(self, docs, state=None): def __call__(self, docs, state=None):
if isinstance(docs, Doc): if isinstance(docs, Doc):
@ -88,6 +91,11 @@ class TokenVectorEncoder(object):
def get_loss(self, docs, golds, scores): def get_loss(self, docs, golds, scores):
raise NotImplementedError raise NotImplementedError
def begin_training(self, gold_tuples, pipeline=None):
self.doc2feats = doc2feats()
if self.model is True:
self.model = self.Model()
class NeuralTagger(object): class NeuralTagger(object):
name = 'nn_tagger' name = 'nn_tagger'
@ -117,15 +125,17 @@ class NeuralTagger(object):
guesses = guesses.get() guesses = guesses.get()
return guesses return guesses
def set_annotations(self, docs, tag_ids): def set_annotations(self, docs, batch_tag_ids):
if isinstance(docs, Doc): if isinstance(docs, Doc):
docs = [docs] docs = [docs]
cdef Doc doc cdef Doc doc
cdef int idx = 0 cdef int idx = 0
cdef int i, j
cdef Vocab vocab = self.vocab
for i, doc in enumerate(docs): for i, doc in enumerate(docs):
tag_ids = tag_ids[idx:idx+len(doc)] doc_tag_ids = batch_tag_ids[idx:idx+len(doc)]
for j, tag_id in enumerate(tag_ids): for j, tag_id in enumerate(doc_tag_ids):
doc.vocab.morphology.assign_tag_id(&doc.c[j], tag_id) vocab.morphology.assign_tag_id(&doc.c[j], tag_id)
idx += 1 idx += 1
def update(self, docs, golds, state=None, drop=0., sgd=None): def update(self, docs, golds, state=None, drop=0., sgd=None):
@ -139,25 +149,19 @@ class NeuralTagger(object):
tag_scores, bp_tag_scores = self.model.begin_update(tokvecs, drop=drop) tag_scores, bp_tag_scores = self.model.begin_update(tokvecs, drop=drop)
loss, d_tag_scores = self.get_loss(docs, golds, tag_scores) loss, d_tag_scores = self.get_loss(docs, golds, tag_scores)
d_tokvecs = bp_tag_scores(d_tag_scores, sgd)
d_tokvecs = bp_tag_scores(d_tag_scores, sgd=sgd)
bp_tokvecs(d_tokvecs, sgd=sgd) bp_tokvecs(d_tokvecs, sgd=sgd)
state['tag_scores'] = tag_scores state['tag_scores'] = tag_scores
state['bp_tag_scores'] = bp_tag_scores
state['d_tag_scores'] = d_tag_scores
state['tag_loss'] = loss state['tag_loss'] = loss
if 'd_tokvecs' in state:
state['d_tokvecs'] += d_tokvecs
else:
state['d_tokvecs'] = d_tokvecs
return state return state
def get_loss(self, docs, golds, scores): def get_loss(self, docs, golds, scores):
tag_index = {tag: i for i, tag in enumerate(docs[0].vocab.morphology.tag_names)} tag_index = {tag: i for i, tag in enumerate(self.vocab.morphology.tag_names)}
idx = 0 cdef int idx = 0
correct = numpy.zeros((scores.shape[0],), dtype='i') correct = numpy.zeros((scores.shape[0],), dtype='i')
for gold in golds: for gold in golds:
for tag in gold.tags: for tag in gold.tags:
@ -165,10 +169,11 @@ class NeuralTagger(object):
idx += 1 idx += 1
correct = self.model.ops.xp.array(correct) correct = self.model.ops.xp.array(correct)
d_scores = scores - to_categorical(correct, nb_classes=scores.shape[1]) d_scores = scores - to_categorical(correct, nb_classes=scores.shape[1])
return (d_scores**2).sum(), d_scores loss = (d_scores**2).sum()
d_scores = self.model.ops.asarray(d_scores)
return loss, d_scores
def begin_training(self, gold_tuples, pipeline=None): def begin_training(self, gold_tuples, pipeline=None):
# Populate tag map, if anything's missing.
tag_map = dict(self.vocab.morphology.tag_map) tag_map = dict(self.vocab.morphology.tag_map)
for raw_text, annots_brackets in gold_tuples: for raw_text, annots_brackets in gold_tuples:
for annots, brackets in annots_brackets: for annots, brackets in annots_brackets:
@ -176,14 +181,12 @@ class NeuralTagger(object):
for tag in tags: for tag in tags:
if tag not in tag_map: if tag not in tag_map:
tag_map[tag] = {POS: X} tag_map[tag] = {POS: X}
cdef Vocab vocab = self.vocab cdef Vocab vocab = self.vocab
vocab.morphology = Morphology(self.vocab.strings, tag_map, vocab.morphology = Morphology(vocab.strings, tag_map,
self.vocab.morphology.lemmatizer) vocab.morphology.lemmatizer)
self.model = Softmax(self.vocab.morphology.n_tags) self.model = Softmax(self.vocab.morphology.n_tags)
cdef class EntityRecognizer(LinearParser): cdef class EntityRecognizer(LinearParser):
""" """
Annotate named entities on Doc objects. Annotate named entities on Doc objects.