forked from songyouwei/ABSA-PyTorch
-
Notifications
You must be signed in to change notification settings - Fork 0
/
data_utils.py
179 lines (152 loc) · 7.26 KB
/
data_utils.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
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
# -*- coding: utf-8 -*-
# file: data_utils.py
# author: songyouwei <[email protected]>
# Copyright (C) 2018. All Rights Reserved.
import os
import pickle
import numpy as np
import torch
from torch.utils.data import Dataset
from pytorch_transformers import BertTokenizer
def build_tokenizer(fnames, max_seq_len, dat_fname):
if os.path.exists(dat_fname):
print('loading tokenizer:', dat_fname)
tokenizer = pickle.load(open(dat_fname, 'rb'))
else:
text = ''
for fname in fnames:
fin = open(fname, 'r', encoding='utf-8', newline='\n', errors='ignore')
lines = fin.readlines()
fin.close()
for i in range(0, len(lines), 3):
text_left, _, text_right = [s.lower().strip() for s in lines[i].partition("$T$")]
aspect = lines[i + 1].lower().strip()
text_raw = text_left + " " + aspect + " " + text_right
text += text_raw + " "
tokenizer = Tokenizer(max_seq_len)
tokenizer.fit_on_text(text)
pickle.dump(tokenizer, open(dat_fname, 'wb'))
return tokenizer
def _load_word_vec(path, word2idx=None):
fin = open(path, 'r', encoding='utf-8', newline='\n', errors='ignore')
word_vec = {}
for line in fin:
tokens = line.rstrip().split()
if word2idx is None or tokens[0] in word2idx.keys():
word_vec[tokens[0]] = np.asarray(tokens[1:], dtype='float32')
return word_vec
def build_embedding_matrix(word2idx, embed_dim, dat_fname):
if os.path.exists(dat_fname):
print('loading embedding_matrix:', dat_fname)
embedding_matrix = pickle.load(open(dat_fname, 'rb'))
else:
print('loading word vectors...')
embedding_matrix = np.zeros((len(word2idx) + 2, embed_dim)) # idx 0 and len(word2idx)+1 are all-zeros
fname = './glove.twitter.27B/glove.twitter.27B.' + str(embed_dim) + 'd.txt' \
if embed_dim != 300 else './glove.42B.300d.txt'
word_vec = _load_word_vec(fname, word2idx=word2idx)
print('building embedding_matrix:', dat_fname)
for word, i in word2idx.items():
vec = word_vec.get(word)
if vec is not None:
# words not found in embedding index will be all-zeros.
embedding_matrix[i] = vec
pickle.dump(embedding_matrix, open(dat_fname, 'wb'))
return embedding_matrix
def pad_and_truncate(sequence, maxlen, dtype='int64', padding='post', truncating='post', value=0):
x = (np.ones(maxlen) * value).astype(dtype)
if truncating == 'pre':
trunc = sequence[-maxlen:]
else:
trunc = sequence[:maxlen]
trunc = np.asarray(trunc, dtype=dtype)
if padding == 'post':
x[:len(trunc)] = trunc
else:
x[-len(trunc):] = trunc
return x
class Tokenizer(object):
def __init__(self, max_seq_len, lower=True):
self.lower = lower
self.max_seq_len = max_seq_len
self.word2idx = {}
self.idx2word = {}
self.idx = 1
def fit_on_text(self, text):
if self.lower:
text = text.lower()
words = text.split()
for word in words:
if word not in self.word2idx:
self.word2idx[word] = self.idx
self.idx2word[self.idx] = word
self.idx += 1
def text_to_sequence(self, text, reverse=False, padding='post', truncating='post'):
if self.lower:
text = text.lower()
words = text.split()
unknownidx = len(self.word2idx)+1
sequence = [self.word2idx[w] if w in self.word2idx else unknownidx for w in words]
if len(sequence) == 0:
sequence = [0]
if reverse:
sequence = sequence[::-1]
return pad_and_truncate(sequence, self.max_seq_len, padding=padding, truncating=truncating)
class Tokenizer4Bert:
def __init__(self, max_seq_len, pretrained_bert_name):
self.tokenizer = BertTokenizer.from_pretrained(pretrained_bert_name)
self.max_seq_len = max_seq_len
def text_to_sequence(self, text, reverse=False, padding='post', truncating='post'):
sequence = self.tokenizer.convert_tokens_to_ids(self.tokenizer.tokenize(text))
if len(sequence) == 0:
sequence = [0]
if reverse:
sequence = sequence[::-1]
return pad_and_truncate(sequence, self.max_seq_len, padding=padding, truncating=truncating)
class ABSADataset(Dataset):
def __init__(self, fname, tokenizer):
fin = open(fname, 'r', encoding='utf-8', newline='\n', errors='ignore')
lines = fin.readlines()
fin.close()
all_data = []
for i in range(0, len(lines), 3):
text_left, _, text_right = [s.lower().strip() for s in lines[i].partition("$T$")]
aspect = lines[i + 1].lower().strip()
polarity = lines[i + 2].strip()
text_raw_indices = tokenizer.text_to_sequence(text_left + " " + aspect + " " + text_right)
text_raw_without_aspect_indices = tokenizer.text_to_sequence(text_left + " " + text_right)
text_left_indices = tokenizer.text_to_sequence(text_left)
text_left_with_aspect_indices = tokenizer.text_to_sequence(text_left + " " + aspect)
text_right_indices = tokenizer.text_to_sequence(text_right, reverse=True)
text_right_with_aspect_indices = tokenizer.text_to_sequence(" " + aspect + " " + text_right, reverse=True)
aspect_indices = tokenizer.text_to_sequence(aspect)
left_context_len = np.sum(text_left_indices != 0)
aspect_len = np.sum(aspect_indices != 0)
aspect_in_text = torch.tensor([left_context_len.item(), (left_context_len + aspect_len - 1).item()])
polarity = int(polarity) + 1
text_bert_indices = tokenizer.text_to_sequence('[CLS] ' + text_left + " " + aspect + " " + text_right + ' [SEP] ' + aspect + " [SEP]")
bert_segments_ids = np.asarray([0] * (np.sum(text_raw_indices != 0) + 2) + [1] * (aspect_len + 1))
bert_segments_ids = pad_and_truncate(bert_segments_ids, tokenizer.max_seq_len)
text_raw_bert_indices = tokenizer.text_to_sequence("[CLS] " + text_left + " " + aspect + " " + text_right + " [SEP]")
aspect_bert_indices = tokenizer.text_to_sequence("[CLS] " + aspect + " [SEP]")
data = {
'text_bert_indices': text_bert_indices,
'bert_segments_ids': bert_segments_ids,
'text_raw_bert_indices': text_raw_bert_indices,
'aspect_bert_indices': aspect_bert_indices,
'text_raw_indices': text_raw_indices,
'text_raw_without_aspect_indices': text_raw_without_aspect_indices,
'text_left_indices': text_left_indices,
'text_left_with_aspect_indices': text_left_with_aspect_indices,
'text_right_indices': text_right_indices,
'text_right_with_aspect_indices': text_right_with_aspect_indices,
'aspect_indices': aspect_indices,
'aspect_in_text': aspect_in_text,
'polarity': polarity,
}
all_data.append(data)
self.data = all_data
def __getitem__(self, index):
return self.data[index]
def __len__(self):
return len(self.data)