explosion/spaCy
1
1
Fork 0
mirror of https://github.com/explosion/spaCy.git synced 2024-11-10 19:57:17 +03:00
Code Issues Packages Projects Releases Wiki Activity

upsampling and batch processing

Browse Source
This commit is contained in:
svlandeg 2019-05-22 23:40:10 +02:00
parent 1a16490d20
commit 97241a3ed7
3 changed files with 157 additions and 151 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
examples/pipeline/wiki_entity_linking/run_el.py
View File

@ -78,8 +78,15 @@ def evaluate(predictions, golds, to_print=True):
fp = 0 fp = 0
fn = 0 fn = 0
corrects = 0
incorrects = 0
for pred, gold in zip(predictions, golds): for pred, gold in zip(predictions, golds):
is_correct = pred == gold is_correct = pred == gold
if is_correct:
corrects += 1
else:
incorrects += 1
if not pred: if not pred:
if not is_correct: # we don't care about tn if not is_correct: # we don't care about tn
fn += 1 fn += 1
@ -98,12 +105,15 @@ def evaluate(predictions, golds, to_print=True):
recall = 100 * tp / (tp + fn + 0.0000001) recall = 100 * tp / (tp + fn + 0.0000001)
fscore = 2 * recall * precision / (recall + precision + 0.0000001) fscore = 2 * recall * precision / (recall + precision + 0.0000001)
accuracy = corrects / (corrects + incorrects)
if to_print: if to_print:
print("precision", round(precision, 1), "%") print("precision", round(precision, 1), "%")
print("recall", round(recall, 1), "%") print("recall", round(recall, 1), "%")
print("Fscore", round(fscore, 1), "%") print("Fscore", round(fscore, 1), "%")
print("Accuracy", round(accuracy, 1), "%")
return precision, recall, fscore return precision, recall, fscore, accuracy
def _prepare_pipeline(nlp, kb): def _prepare_pipeline(nlp, kb):

294
examples/pipeline/wiki_entity_linking/train_el.py
View File

@ -6,6 +6,7 @@ import datetime
from os import listdir from os import listdir
import numpy as np import numpy as np
import random import random
from random import shuffle
from thinc.neural._classes.convolution import ExtractWindow from thinc.neural._classes.convolution import ExtractWindow
from examples.pipeline.wiki_entity_linking import run_el, training_set_creator, kb_creator from examples.pipeline.wiki_entity_linking import run_el, training_set_creator, kb_creator
@ -26,17 +27,17 @@ from spacy.tokens import Doc
class EL_Model: class EL_Model:
PRINT_LOSS = False
PRINT_F = True
PRINT_TRAIN = False PRINT_TRAIN = False
EPS = 0.0000000005 EPS = 0.0000000005
CUTOFF = 0.5 CUTOFF = 0.5
BATCH_SIZE = 5
INPUT_DIM = 300 INPUT_DIM = 300
HIDDEN_1_WIDTH = 256 # 10 HIDDEN_1_WIDTH = 32 # 10
HIDDEN_2_WIDTH = 32 # 6 HIDDEN_2_WIDTH = 32 # 6
ENTITY_WIDTH = 64 # 4 DESC_WIDTH = 64 # 4
ARTICLE_WIDTH = 128 # 8 ARTICLE_WIDTH = 64 # 8
DROP = 0.1 DROP = 0.1
@ -48,7 +49,7 @@ class EL_Model:
self.kb = kb self.kb = kb
self._build_cnn(in_width=self.INPUT_DIM, self._build_cnn(in_width=self.INPUT_DIM,
entity_width=self.ENTITY_WIDTH, desc_width=self.DESC_WIDTH,
article_width=self.ARTICLE_WIDTH, article_width=self.ARTICLE_WIDTH,
hidden_1_width=self.HIDDEN_1_WIDTH, hidden_1_width=self.HIDDEN_1_WIDTH,
hidden_2_width=self.HIDDEN_2_WIDTH) hidden_2_width=self.HIDDEN_2_WIDTH)
@ -57,121 +58,118 @@ class EL_Model:
# raise errors instead of runtime warnings in case of int/float overflow # raise errors instead of runtime warnings in case of int/float overflow
np.seterr(all='raise') np.seterr(all='raise')
train_inst, train_pos, train_neg, train_texts = self._get_training_data(training_dir, train_ent, train_gold, train_desc, train_article, train_texts = self._get_training_data(training_dir,
entity_descr_output, entity_descr_output,
False, False,
trainlimit, trainlimit,
balance=True, to_print=False)
to_print=False)
train_pos_entities = [k for k,v in train_gold.items() if v]
train_neg_entities = [k for k,v in train_gold.items() if not v]
train_pos_count = len(train_pos_entities)
train_neg_count = len(train_neg_entities)
# upsample positives to 50-50 distribution
while train_pos_count < train_neg_count:
train_ent.append(random.choice(train_pos_entities))
train_pos_count += 1
# upsample negatives to 50-50 distribution
while train_neg_count < train_pos_count:
train_ent.append(random.choice(train_neg_entities))
train_neg_count += 1
shuffle(train_ent)
dev_ent, dev_gold, dev_desc, dev_article, dev_texts = self._get_training_data(training_dir,
entity_descr_output,
True,
devlimit,
to_print=False)
shuffle(dev_ent)
dev_pos_count = len([g for g in dev_gold.values() if g])
dev_neg_count = len([g for g in dev_gold.values() if not g])
dev_inst, dev_pos, dev_neg, dev_texts = self._get_training_data(training_dir,
entity_descr_output,
True,
devlimit,
balance=False,
to_print=False)
self._begin_training() self._begin_training()
print() print()
self._test_dev(dev_inst, dev_pos, dev_neg, dev_texts, print_string="dev_random", calc_random=True) self._test_dev(dev_ent, dev_gold, dev_desc, dev_article, dev_texts, print_string="dev_random", calc_random=True)
self._test_dev(dev_inst, dev_pos, dev_neg, dev_texts, print_string="dev_pre", avg=False) print()
self._test_dev(dev_ent, dev_gold, dev_desc, dev_article, dev_texts, print_string="dev_pre", avg=True)
instance_pos_count = 0
instance_neg_count = 0
if to_print: if to_print:
print() print()
print("Training on", len(train_inst.values()), "articles") print("Training on", len(train_ent), "entities in", len(train_texts), "articles")
print("Dev test on", len(dev_inst.values()), "articles") print("Training instances pos/neg", train_pos_count, train_neg_count)
print()
print("Dev test on", len(dev_ent), "entities in", len(dev_texts), "articles")
print("Dev instances pos/neg", dev_pos_count, dev_neg_count)
print() print()
print(" CUTOFF", self.CUTOFF) print(" CUTOFF", self.CUTOFF)
print(" INPUT_DIM", self.INPUT_DIM) print(" INPUT_DIM", self.INPUT_DIM)
print(" HIDDEN_1_WIDTH", self.HIDDEN_1_WIDTH) print(" HIDDEN_1_WIDTH", self.HIDDEN_1_WIDTH)
print(" ENTITY_WIDTH", self.ENTITY_WIDTH) print(" DESC_WIDTH", self.DESC_WIDTH)
print(" ARTICLE_WIDTH", self.ARTICLE_WIDTH) print(" ARTICLE_WIDTH", self.ARTICLE_WIDTH)
print(" HIDDEN_2_WIDTH", self.HIDDEN_2_WIDTH) print(" HIDDEN_2_WIDTH", self.HIDDEN_2_WIDTH)
print(" DROP", self.DROP) print(" DROP", self.DROP)
print() print()
# TODO: proper batches. Currently 1 article at the time start = 0
# TODO shuffle data (currently positive is always followed by several negatives) stop = min(self.BATCH_SIZE, len(train_ent))
article_count = 0 processed = 0
for article_id, inst_cluster_set in train_inst.items():
try:
# if to_print:
# print()
# print(article_count, "Training on article", article_id)
article_count += 1
article_text = train_texts[article_id]
entities = list()
golds = list()
for inst_cluster in inst_cluster_set:
entities.append(train_pos.get(inst_cluster))
golds.append(float(1.0))
instance_pos_count += 1
for neg_entity in train_neg.get(inst_cluster, []):
entities.append(neg_entity)
golds.append(float(0.0))
instance_neg_count += 1
self.update(article_text=article_text, entities=entities, golds=golds) while start < len(train_ent):
next_batch = train_ent[start:stop]
# dev eval golds = [train_gold[e] for e in next_batch]
self._test_dev(dev_inst, dev_pos, dev_neg, dev_texts, print_string="dev_inter_avg", avg=True) descs = [train_desc[e] for e in next_batch]
except ValueError as e: articles = [train_texts[train_article[e]] for e in next_batch]
print("Error in article id", article_id)
self.update(entities=next_batch, golds=golds, descs=descs, texts=articles)
self._test_dev(dev_ent, dev_gold, dev_desc, dev_article, dev_texts, print_string="dev_inter", avg=True)
processed += len(next_batch)
start = start + self.BATCH_SIZE
stop = min(stop + self.BATCH_SIZE, len(train_ent))
if to_print: if to_print:
print() print()
print("Trained on", instance_pos_count, "/", instance_neg_count, "instances pos/neg") print("Trained on", processed, "entities in total")
def _test_dev(self, instances, pos, neg, texts_by_id, print_string, avg=False, calc_random=False): def _test_dev(self, entities, gold_by_entity, desc_by_entity, article_by_entity, texts_by_id, print_string, avg=True, calc_random=False):
predictions = list() golds = [gold_by_entity[e] for e in entities]
golds = list()
for article_id, inst_cluster_set in instances.items(): if calc_random:
for inst_cluster in inst_cluster_set: predictions = self._predict_random(entities=entities)
pos_ex = pos.get(inst_cluster)
neg_exs = neg.get(inst_cluster, [])
article = inst_cluster.split(sep="_")[0] else:
entity_id = inst_cluster.split(sep="_")[1] desc_docs = self.nlp.pipe([desc_by_entity[e] for e in entities])
article_doc = self.nlp(texts_by_id[article]) article_docs = self.nlp.pipe([texts_by_id[article_by_entity[e]] for e in entities])
entities = [self.nlp(pos_ex)] predictions = self._predict(entities=entities, article_docs=article_docs, desc_docs=desc_docs, avg=avg)
golds.append(float(1.0))
for neg_ex in neg_exs:
entities.append(self.nlp(neg_ex))
golds.append(float(0.0))
if calc_random:
preds = self._predict_random(entities=entities)
else:
preds = self._predict(article_doc=article_doc, entities=entities, avg=avg)
predictions.extend(preds)
# TODO: combine with prior probability # TODO: combine with prior probability
p, r, f = run_el.evaluate(predictions, golds, to_print=False) p, r, f, acc = run_el.evaluate(predictions, golds, to_print=False)
if self.PRINT_F:
print("p/r/F", print_string, round(p, 1), round(r, 1), round(f, 1))
loss, gradient = self.get_loss(self.model.ops.asarray(predictions), self.model.ops.asarray(golds)) loss, gradient = self.get_loss(self.model.ops.asarray(predictions), self.model.ops.asarray(golds))
if self.PRINT_LOSS:
print("loss", print_string, round(loss, 5)) print("p/r/F/acc/loss", print_string, round(p, 1), round(r, 1), round(f, 1), round(acc, 2), round(loss, 5))
return loss, p, r, f return loss, p, r, f
def _predict(self, article_doc, entities, avg=False, apply_threshold=True): def _predict(self, entities, article_docs, desc_docs, avg=True, apply_threshold=True):
if avg: if avg:
with self.article_encoder.use_params(self.sgd_article.averages) \ with self.article_encoder.use_params(self.sgd_article.averages) \
and self.entity_encoder.use_params(self.sgd_entity.averages): and self.desc_encoder.use_params(self.sgd_entity.averages):
doc_encoding = self.article_encoder([article_doc])[0] doc_encodings = self.article_encoder(article_docs)
entity_encodings = self.entity_encoder(entities) desc_encodings = self.desc_encoder(desc_docs)
else: else:
doc_encoding = self.article_encoder([article_doc])[0] doc_encodings = self.article_encoder(article_docs)
entity_encodings = self.entity_encoder(entities) desc_encodings = self.desc_encoder(desc_docs)
concat_encodings = [list(entity_encodings[i]) + list(doc_encoding) for i in range(len(entities))] concat_encodings = [list(desc_encodings[i]) + list(doc_encodings[i]) for i in range(len(entities))]
np_array_list = np.asarray(concat_encodings) np_array_list = np.asarray(concat_encodings)
if avg: if avg:
@ -189,16 +187,16 @@ class EL_Model:
def _predict_random(self, entities, apply_threshold=True): def _predict_random(self, entities, apply_threshold=True):
if not apply_threshold: if not apply_threshold:
return [float(random.uniform(0,1)) for e in entities] return [float(random.uniform(0, 1)) for e in entities]
else: else:
return [float(1.0) if random.uniform(0,1) > self.CUTOFF else float(0.0) for e in entities] return [float(1.0) if random.uniform(0, 1) > self.CUTOFF else float(0.0) for e in entities]
def _build_cnn(self, in_width, entity_width, article_width, hidden_1_width, hidden_2_width): def _build_cnn(self, in_width, desc_width, article_width, hidden_1_width, hidden_2_width):
with Model.define_operators({">>": chain, "|": concatenate, "**": clone}): with Model.define_operators({">>": chain, "|": concatenate, "**": clone}):
self.entity_encoder = self._encoder(in_width=in_width, hidden_with=hidden_1_width, end_width=entity_width) self.desc_encoder = self._encoder(in_width=in_width, hidden_with=hidden_1_width, end_width=desc_width)
self.article_encoder = self._encoder(in_width=in_width, hidden_with=hidden_1_width, end_width=article_width) self.article_encoder = self._encoder(in_width=in_width, hidden_with=hidden_1_width, end_width=article_width)
in_width = entity_width + article_width in_width = desc_width + article_width
out_width = hidden_2_width out_width = hidden_2_width
self.model = Affine(out_width, in_width) \ self.model = Affine(out_width, in_width) \
@ -229,80 +227,78 @@ class EL_Model:
def _begin_training(self): def _begin_training(self):
self.sgd_article = create_default_optimizer(self.article_encoder.ops) self.sgd_article = create_default_optimizer(self.article_encoder.ops)
self.sgd_entity = create_default_optimizer(self.entity_encoder.ops) self.sgd_entity = create_default_optimizer(self.desc_encoder.ops)
self.sgd = create_default_optimizer(self.model.ops) self.sgd = create_default_optimizer(self.model.ops)
@staticmethod @staticmethod
def get_loss(predictions, golds): def get_loss(predictions, golds):
d_scores = (predictions - golds) d_scores = (predictions - golds)
gradient = d_scores.mean()
loss = (d_scores ** 2).mean() loss = (d_scores ** 2).mean()
return loss, d_scores return loss, gradient
# TODO: multiple docs/articles def update(self, entities, golds, descs, texts):
def update(self, article_text, entities, golds, apply_threshold=True): golds = self.model.ops.asarray(golds)
article_doc = self.nlp(article_text)
# entity_docs = list(self.nlp.pipe(entities))
for entity, gold in zip(entities, golds): desc_docs = self.nlp.pipe(descs)
doc_encodings, bp_doc = self.article_encoder.begin_update([article_doc], drop=self.DROP) article_docs = self.nlp.pipe(texts)
doc_encoding = doc_encodings[0]
entity_doc = self.nlp(entity) doc_encodings, bp_doc = self.article_encoder.begin_update(article_docs, drop=self.DROP)
# print("entity_docs", type(entity_doc))
entity_encodings, bp_entity = self.entity_encoder.begin_update([entity_doc], drop=self.DROP) desc_encodings, bp_entity = self.desc_encoder.begin_update(desc_docs, drop=self.DROP)
entity_encoding = entity_encodings[0]
# print("entity_encoding", len(entity_encoding), entity_encoding)
concat_encodings = [list(entity_encoding) + list(doc_encoding)] # for i in range(len(entities)) concat_encodings = [list(desc_encodings[i]) + list(doc_encodings[i]) for i in range(len(entities))]
# print("concat_encodings", len(concat_encodings), concat_encodings)
prediction, bp_model = self.model.begin_update(np.asarray(concat_encodings), drop=self.DROP) predictions, bp_model = self.model.begin_update(np.asarray(concat_encodings), drop=self.DROP)
# predictions = self.model.ops.flatten(predictions) predictions = self.model.ops.flatten(predictions)
# print("prediction", prediction) # print("entities", entities)
# golds = self.model.ops.asarray(golds) # print("predictions", predictions)
# print("gold", gold) # print("golds", golds)
loss, gradient = self.get_loss(prediction, gold) loss, gradient = self.get_loss(predictions, golds)
if self.PRINT_LOSS and self.PRINT_TRAIN: if self.PRINT_TRAIN:
print("loss train", round(loss, 5)) print("loss train", round(loss, 5))
gradient = float(gradient) gradient = float(gradient)
# print("gradient", gradient) # print("gradient", gradient)
# print("loss", loss) # print("loss", loss)
model_gradient = bp_model(gradient, sgd=self.sgd) model_gradient = bp_model(gradient, sgd=self.sgd)
# print("model_gradient", model_gradient) # print("model_gradient", model_gradient)
# concat = entity + doc, but doc is the same within this function (TODO: multiple docs/articles) # concat = desc + doc, but doc is the same within this function (TODO: multiple docs/articles)
doc_gradient = model_gradient[0][self.ENTITY_WIDTH:] doc_gradient = model_gradient[0][self.DESC_WIDTH:]
entity_gradients = list() entity_gradients = list()
for x in model_gradient: for x in model_gradient:
entity_gradients.append(list(x[0:self.ENTITY_WIDTH])) entity_gradients.append(list(x[0:self.DESC_WIDTH]))
# print("doc_gradient", doc_gradient) # print("doc_gradient", doc_gradient)
# print("entity_gradients", entity_gradients) # print("entity_gradients", entity_gradients)
bp_doc([doc_gradient], sgd=self.sgd_article) bp_doc([doc_gradient], sgd=self.sgd_article)
bp_entity(entity_gradients, sgd=self.sgd_entity) bp_entity(entity_gradients, sgd=self.sgd_entity)
def _get_training_data(self, training_dir, entity_descr_output, dev, limit, balance, to_print): def _get_training_data(self, training_dir, entity_descr_output, dev, limit, to_print):
id_to_descr = kb_creator._get_id_to_description(entity_descr_output) id_to_descr = kb_creator._get_id_to_description(entity_descr_output)
correct_entries, incorrect_entries = training_set_creator.read_training_entities(training_output=training_dir, correct_entries, incorrect_entries = training_set_creator.read_training_entities(training_output=training_dir,
collect_correct=True, collect_correct=True,
collect_incorrect=True) collect_incorrect=True)
instance_by_article = dict()
local_vectors = list() # TODO: local vectors local_vectors = list() # TODO: local vectors
text_by_article = dict() text_by_article = dict()
pos_entities = dict() gold_by_entity = dict()
neg_entities = dict() desc_by_entity = dict()
article_by_entity = dict()
entities = list()
cnt = 0 cnt = 0
for f in listdir(training_dir): next_entity_nr = 0
files = listdir(training_dir)
shuffle(files)
for f in files:
if not limit or cnt < limit: if not limit or cnt < limit:
if dev == run_el.is_dev(f): if dev == run_el.is_dev(f):
article_id = f.replace(".txt", "") article_id = f.replace(".txt", "")
@ -313,29 +309,29 @@ class EL_Model:
with open(os.path.join(training_dir, f), mode="r", encoding='utf8') as file: with open(os.path.join(training_dir, f), mode="r", encoding='utf8') as file:
text = file.read() text = file.read()
text_by_article[article_id] = text text_by_article[article_id] = text
instance_by_article[article_id] = set()
for mention, entity_pos in correct_entries[article_id].items(): for mention, entity_pos in correct_entries[article_id].items():
descr = id_to_descr.get(entity_pos) descr = id_to_descr.get(entity_pos)
if descr: if descr:
instance_by_article[article_id].add(article_id + "_" + mention) entities.append(next_entity_nr)
pos_entities[article_id + "_" + mention] = descr gold_by_entity[next_entity_nr] = 1
desc_by_entity[next_entity_nr] = descr
article_by_entity[next_entity_nr] = article_id
next_entity_nr += 1
for mention, entity_negs in incorrect_entries[article_id].items(): for mention, entity_negs in incorrect_entries[article_id].items():
if not balance or pos_entities.get(article_id + "_" + mention): for entity_neg in entity_negs:
neg_count = 0 descr = id_to_descr.get(entity_neg)
for entity_neg in entity_negs: if descr:
# if balance, keep only 1 negative instance for each positive instance entities.append(next_entity_nr)
if neg_count < 1 or not balance: gold_by_entity[next_entity_nr] = 0
descr = id_to_descr.get(entity_neg) desc_by_entity[next_entity_nr] = descr
if descr: article_by_entity[next_entity_nr] = article_id
descr_list = neg_entities.get(article_id + "_" + mention, []) next_entity_nr += 1
descr_list.append(descr)
neg_entities[article_id + "_" + mention] = descr_list
neg_count += 1
if to_print: if to_print:
print() print()
print("Processed", cnt, "training articles, dev=" + str(dev)) print("Processed", cnt, "training articles, dev=" + str(dev))
print() print()
return instance_by_article, pos_entities, neg_entities, text_by_article return entities, gold_by_entity, desc_by_entity, article_by_entity, text_by_article

2
examples/pipeline/wiki_entity_linking/wiki_nel_pipeline.py
View File

@ -111,7 +111,7 @@ if __name__ == "__main__":
print("STEP 6: training", datetime.datetime.now()) print("STEP 6: training", datetime.datetime.now())
my_nlp = spacy.load('en_core_web_md') my_nlp = spacy.load('en_core_web_md')
trainer = EL_Model(kb=my_kb, nlp=my_nlp) trainer = EL_Model(kb=my_kb, nlp=my_nlp)
trainer.train_model(training_dir=TRAINING_DIR, entity_descr_output=ENTITY_DESCR, trainlimit=100, devlimit=20) trainer.train_model(training_dir=TRAINING_DIR, entity_descr_output=ENTITY_DESCR, trainlimit=400, devlimit=50)
print() print()
# STEP 7: apply the EL algorithm on the dev dataset # STEP 7: apply the EL algorithm on the dev dataset