-
Notifications
You must be signed in to change notification settings - Fork 116
/
embeddings.py
150 lines (124 loc) · 4.57 KB
/
embeddings.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
# -*- coding: utf-8 -*-
"""
Load word embeddings from different representations.
"""
from __future__ import print_function
import os
import numpy as np
import logging
from itertools import izip
# local
from word_dictionary import WordDictionary
# ----------------------------------------------------------------------
class Plain(object):
@classmethod
def read_vectors(cls, filename):
"""
Read an embedding from a plain text file with one vector per
line, values separated by whitespace.
"""
with open(filename, 'rb') as file:
matrix = np.array([[float(value) for value in line.split()]
for line in file])
return matrix
@classmethod
def read_vocabulary(cls, filename):
"""
Read a vocabulary file containing one word per line.
Return a list of words.
"""
words = []
with open(filename, 'rb') as f:
for line in f:
word = unicode(line.strip(), 'utf-8')
if word:
words.append(word)
return words
@classmethod
def write_vocabulary(cls, vocab, filename):
"""
Write a vocabulary to a file containing one word per line.
"""
with open(filename, 'wb') as f:
for word in vocab:
print(word.encode('utf-8'), file=f)
@classmethod
def write_vectors(cls, filename, matrix):
"""
Write embedding vectors to a plain text file with one vector per
line, values separated by whitespace.
"""
with open(filename, 'wb') as file:
for row in matrix:
print(' '.join(["%f" % x for x in row]), file=file)
# ----------------------------------------------------------------------
class Senna(object):
@classmethod
def read_vocabulary(cls, filename):
"""
Read the vocabulary file used by SENNA.
It has one word per line, all lower case except for the special words
PADDING and UNKNOWN.
"""
return Plain.vocabulary(filename)
# ----------------------------------------------------------------------
class Word2Embeddings(object):
@classmethod
def read_vocabulary(cls, filename):
"""
Read the vocabulary used with word2embeddings.
It is the same as a plain text vocabulary, except the embeddings for
the rare/unknown word are the first two items (before any word in the file).
"""
return Plain.vocabulary(filename, 'polyglot')
@classmethod
def read_vectors(cls, filename):
"""
Load the feature matrix used by word2embeddings.
"""
import cPickle as pickle
with open(filename, 'rb') as f:
model = pickle.load(f)
return model.get_word_embeddings()
# ----------------------------------------------------------------------
class Word2Vec(object):
@classmethod
def load(cls, filename):
"""
Load words and vectors from a file in word2vec format.
"""
words = []
vectors = []
with open(filename, 'rb') as f:
len, size = f.readline().split()
for line in f:
items = line.split()
word = unicode(items[0], 'utf-8')
words.append(word)
vectors.append([float(x) for x in items[1:]])
# vectors for the special symbols, not present in words, will be
# created later
return np.array(vectors), words
@classmethod
def save(cls, filename, words, vectors):
"""
Save words and vectors to a file in word2vec format.
:param vectors: is a Numpy array
"""
with open(filename, 'wb') as f:
print(len(words), vectors.shape[1], file=f)
for word, vector in izip(words, vectors):
print(word.encode('UTF-8'), ' '.join('%f' % w for w in vector), file=f)
# ----------------------------------------------------------------------
def generate_vectors(num_vectors, num_features, min_value=-0.1, max_value=0.1):
"""
Generates vectors of real numbers, to be used as word features.
Vectors are initialized randomly with values in the interval [min_value, max_value]
:return: a 2-dim numpy array.
"""
# set the seed for replicability
#np.random.seed(42) # DEBUG
table = np.random.uniform(min_value, max_value, (num_vectors, num_features))
logging.debug("Generated %d feature vectors with %d features each." %
(num_vectors, num_features))
return table