2019-06-05 01:09:46 +03:00
|
|
|
from random import shuffle
|
|
|
|
|
|
|
|
from examples.pipeline.wiki_entity_linking import kb_creator
|
|
|
|
|
|
|
|
import numpy as np
|
|
|
|
|
|
|
|
from spacy._ml import zero_init, create_default_optimizer
|
|
|
|
from spacy.cli.pretrain import get_cossim_loss
|
|
|
|
|
|
|
|
from thinc.v2v import Model
|
|
|
|
from thinc.api import chain
|
|
|
|
from thinc.neural._classes.affine import Affine
|
|
|
|
|
|
|
|
|
|
|
|
class EntityEncoder:
|
|
|
|
|
|
|
|
DROP = 0
|
|
|
|
EPOCHS = 5
|
2019-06-06 20:51:27 +03:00
|
|
|
STOP_THRESHOLD = 0.9 # 0.1
|
2019-06-05 01:09:46 +03:00
|
|
|
|
|
|
|
BATCH_SIZE = 1000
|
|
|
|
|
2019-06-06 20:51:27 +03:00
|
|
|
def __init__(self, nlp, input_dim, desc_width):
|
2019-06-05 01:09:46 +03:00
|
|
|
self.nlp = nlp
|
2019-06-06 20:51:27 +03:00
|
|
|
self.input_dim = input_dim
|
|
|
|
self.desc_width = desc_width
|
|
|
|
|
|
|
|
def apply_encoder(self, description_list):
|
|
|
|
if self.encoder is None:
|
|
|
|
raise ValueError("Can not apply encoder before training it")
|
|
|
|
|
|
|
|
print("Encoding", len(description_list), "entities")
|
|
|
|
|
|
|
|
batch_size = 10000
|
2019-06-05 01:09:46 +03:00
|
|
|
|
2019-06-06 20:51:27 +03:00
|
|
|
start = 0
|
|
|
|
stop = min(batch_size, len(description_list))
|
|
|
|
encodings = []
|
2019-06-05 01:09:46 +03:00
|
|
|
|
2019-06-06 20:51:27 +03:00
|
|
|
while start < len(description_list):
|
|
|
|
docs = list(self.nlp.pipe(description_list[start:stop]))
|
|
|
|
doc_embeddings = [self._get_doc_embedding(doc) for doc in docs]
|
|
|
|
enc = self.encoder(np.asarray(doc_embeddings))
|
|
|
|
encodings.extend(enc.tolist())
|
2019-06-05 01:09:46 +03:00
|
|
|
|
2019-06-06 20:51:27 +03:00
|
|
|
start = start + batch_size
|
|
|
|
stop = min(stop + batch_size, len(description_list))
|
|
|
|
print("encoded :", len(encodings))
|
2019-06-05 01:09:46 +03:00
|
|
|
|
2019-06-06 20:51:27 +03:00
|
|
|
return encodings
|
2019-06-05 19:29:18 +03:00
|
|
|
|
2019-06-06 20:51:27 +03:00
|
|
|
def train(self, description_list, to_print=False):
|
|
|
|
processed, loss = self._train_model(description_list)
|
|
|
|
|
|
|
|
if to_print:
|
|
|
|
print("Trained on", processed, "entities across", self.EPOCHS, "epochs")
|
|
|
|
print("Final loss:", loss)
|
|
|
|
|
|
|
|
# self._test_encoder()
|
|
|
|
|
|
|
|
def _train_model(self, description_list):
|
2019-06-05 01:09:46 +03:00
|
|
|
# TODO: when loss gets too low, a 'mean of empty slice' warning is thrown by numpy
|
|
|
|
|
2019-06-06 20:51:27 +03:00
|
|
|
self._build_network(self.input_dim, self.desc_width)
|
2019-06-05 01:09:46 +03:00
|
|
|
|
|
|
|
processed = 0
|
|
|
|
loss = 1
|
2019-06-06 20:51:27 +03:00
|
|
|
descriptions = description_list.copy() # copy this list so that shuffling does not affect other functions
|
2019-06-05 01:09:46 +03:00
|
|
|
|
|
|
|
for i in range(self.EPOCHS):
|
2019-06-06 20:51:27 +03:00
|
|
|
shuffle(descriptions)
|
2019-06-05 01:09:46 +03:00
|
|
|
|
|
|
|
batch_nr = 0
|
|
|
|
start = 0
|
2019-06-06 20:51:27 +03:00
|
|
|
stop = min(self.BATCH_SIZE, len(descriptions))
|
2019-06-05 01:09:46 +03:00
|
|
|
|
2019-06-06 20:51:27 +03:00
|
|
|
while loss > self.STOP_THRESHOLD and start < len(descriptions):
|
2019-06-05 01:09:46 +03:00
|
|
|
batch = []
|
2019-06-06 20:51:27 +03:00
|
|
|
for descr in descriptions[start:stop]:
|
2019-06-05 01:09:46 +03:00
|
|
|
doc = self.nlp(descr)
|
|
|
|
doc_vector = self._get_doc_embedding(doc)
|
|
|
|
batch.append(doc_vector)
|
|
|
|
|
2019-06-06 20:51:27 +03:00
|
|
|
loss = self._update(batch)
|
2019-06-05 01:09:46 +03:00
|
|
|
print(i, batch_nr, loss)
|
|
|
|
processed += len(batch)
|
|
|
|
|
|
|
|
batch_nr += 1
|
|
|
|
start = start + self.BATCH_SIZE
|
2019-06-06 20:51:27 +03:00
|
|
|
stop = min(stop + self.BATCH_SIZE, len(descriptions))
|
2019-06-05 01:09:46 +03:00
|
|
|
|
|
|
|
return processed, loss
|
|
|
|
|
|
|
|
@staticmethod
|
|
|
|
def _get_doc_embedding(doc):
|
|
|
|
indices = np.zeros((len(doc),), dtype="i")
|
|
|
|
for i, word in enumerate(doc):
|
|
|
|
if word.orth in doc.vocab.vectors.key2row:
|
|
|
|
indices[i] = doc.vocab.vectors.key2row[word.orth]
|
|
|
|
else:
|
|
|
|
indices[i] = 0
|
|
|
|
word_vectors = doc.vocab.vectors.data[indices]
|
|
|
|
doc_vector = np.mean(word_vectors, axis=0) # TODO: min? max?
|
|
|
|
return doc_vector
|
|
|
|
|
|
|
|
def _build_network(self, orig_width, hidden_with):
|
|
|
|
with Model.define_operators({">>": chain}):
|
|
|
|
self.encoder = (
|
|
|
|
Affine(hidden_with, orig_width)
|
|
|
|
)
|
|
|
|
self.model = self.encoder >> zero_init(Affine(orig_width, hidden_with, drop_factor=0.0))
|
|
|
|
|
|
|
|
self.sgd = create_default_optimizer(self.model.ops)
|
|
|
|
|
2019-06-06 20:51:27 +03:00
|
|
|
def _update(self, vectors):
|
2019-06-05 01:09:46 +03:00
|
|
|
predictions, bp_model = self.model.begin_update(np.asarray(vectors), drop=self.DROP)
|
|
|
|
|
2019-06-06 20:51:27 +03:00
|
|
|
loss, d_scores = self._get_loss(scores=predictions, golds=np.asarray(vectors))
|
2019-06-05 01:09:46 +03:00
|
|
|
bp_model(d_scores, sgd=self.sgd)
|
|
|
|
|
|
|
|
return loss / len(vectors)
|
|
|
|
|
|
|
|
@staticmethod
|
2019-06-06 20:51:27 +03:00
|
|
|
def _get_loss(golds, scores):
|
2019-06-05 01:09:46 +03:00
|
|
|
loss, gradients = get_cossim_loss(scores, golds)
|
|
|
|
return loss, gradients
|
2019-06-05 19:29:18 +03:00
|
|
|
|
|
|
|
def _test_encoder(self):
|
|
|
|
""" Test encoder on some dummy examples """
|
|
|
|
desc_A1 = "Fictional character in The Simpsons"
|
|
|
|
desc_A2 = "Simpsons - fictional human"
|
|
|
|
desc_A3 = "Fictional character in The Flintstones"
|
|
|
|
desc_A4 = "Politician from the US"
|
|
|
|
|
|
|
|
A1_doc_vector = np.asarray([self._get_doc_embedding(self.nlp(desc_A1))])
|
|
|
|
A2_doc_vector = np.asarray([self._get_doc_embedding(self.nlp(desc_A2))])
|
|
|
|
A3_doc_vector = np.asarray([self._get_doc_embedding(self.nlp(desc_A3))])
|
|
|
|
A4_doc_vector = np.asarray([self._get_doc_embedding(self.nlp(desc_A4))])
|
|
|
|
|
|
|
|
loss_a1_a1, _ = get_cossim_loss(A1_doc_vector, A1_doc_vector)
|
|
|
|
loss_a1_a2, _ = get_cossim_loss(A1_doc_vector, A2_doc_vector)
|
|
|
|
loss_a1_a3, _ = get_cossim_loss(A1_doc_vector, A3_doc_vector)
|
|
|
|
loss_a1_a4, _ = get_cossim_loss(A1_doc_vector, A4_doc_vector)
|
|
|
|
|
|
|
|
print("sim doc A1 A1", loss_a1_a1)
|
|
|
|
print("sim doc A1 A2", loss_a1_a2)
|
|
|
|
print("sim doc A1 A3", loss_a1_a3)
|
|
|
|
print("sim doc A1 A4", loss_a1_a4)
|
|
|
|
|
|
|
|
A1_encoded = self.encoder(A1_doc_vector)
|
|
|
|
A2_encoded = self.encoder(A2_doc_vector)
|
|
|
|
A3_encoded = self.encoder(A3_doc_vector)
|
|
|
|
A4_encoded = self.encoder(A4_doc_vector)
|
|
|
|
|
|
|
|
loss_a1_a1, _ = get_cossim_loss(A1_encoded, A1_encoded)
|
|
|
|
loss_a1_a2, _ = get_cossim_loss(A1_encoded, A2_encoded)
|
|
|
|
loss_a1_a3, _ = get_cossim_loss(A1_encoded, A3_encoded)
|
|
|
|
loss_a1_a4, _ = get_cossim_loss(A1_encoded, A4_encoded)
|
|
|
|
|
|
|
|
print("sim encoded A1 A1", loss_a1_a1)
|
|
|
|
print("sim encoded A1 A2", loss_a1_a2)
|
|
|
|
print("sim encoded A1 A3", loss_a1_a3)
|
|
|
|
print("sim encoded A1 A4", loss_a1_a4)
|