-
Notifications
You must be signed in to change notification settings - Fork 0
/
preprocess.py
173 lines (144 loc) · 6.38 KB
/
preprocess.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
import sys, pickle
import tensorflow as tf
import numpy as np
import pandas as pd
from data_helper import Vocabulary
from argparser import model_opts
sys.path.append("..")
def load_data(args):
vocab = Vocabulary()
vocab.load(args.vocab_file, keep_words=args.vocab_size)
df_train = pd.read_csv(args.train_data_file, sep="\t")
df_train.fillna(value="", inplace=True)
print("train:", df_train.shape)
df_dev = pd.read_csv(args.dev_data_file, sep="\t")
df_dev.fillna(value="", inplace=True)
print("dev:", df_dev.shape)
df_test = pd.read_csv(args.test_data_file, sep="\t")
df_test.fillna(value="", inplace=True)
print("test:", df_test.shape)
df_train_sim = pd.read_csv(args.train_sim_file, sep="\t")
df_train_sim.fillna(value="", inplace=True)
print("train_sim:", df_train_sim.shape)
df_dev_sim = pd.read_csv(args.dev_sim_file, sep="\t")
df_dev_sim.fillna(value="", inplace=True)
print("dev_sim:", df_dev_sim.shape)
df_test_sim = pd.read_csv(args.test_sim_file, sep="\t")
df_test_sim.fillna(value="", inplace=True)
print("test_sim:", df_test_sim.shape)
def _do_vectorize(df, name):
df = df.copy()
df["sentence"] = df["sentence"].map(eval)
grouped = df.groupby("doc")
sentence_nums = []
sentence_cut_words = []
sentence_word_ids = []
sentences_lens = []
roles = []
zero = pd.Series([0])
for agg_name, agg_df in grouped:
if args.padding_data:
sentence_nums.append(args.max_sentence_num)
role = agg_df["role"]
if len(role) >= args.max_sentence_num:
roles.append(role[:args.max_sentence_num])
else:
for i in range(args.max_sentence_num - len(agg_df["role"])):
role = role.append(zero)
roles.append(role)
else:
if args.intercept and name:
if len(agg_df) >= args.max_sentence_num:
sentence_nums.append(args.max_sentence_num)
roles.append(agg_df["role"][-args.max_sentence_num:])
else:
sentence_nums.append(len(agg_df))
roles.append(agg_df["role"])
else:
sentence_nums.append(len(agg_df))
roles.append(agg_df["role"])
tmp_words = []
i = 0
for words in agg_df["sentence"]:
i += 1
if len(words) <= args.max_sequence_length:
tmp_words.append(words)
else:
tmp_words.append(words[:args.max_sequence_length])
if args.padding_data:
if i == args.max_sentence_num:
break
if args.intercept and name:
if len(tmp_words) > args.max_sentence_num:
tmp_words = tmp_words[-args.max_sentence_num:]
if args.padding_data:
# sentences_lens.append([args.max_sequence_length for x in tmp_words])
sentences_lens.append([args.max_sequence_length for i in range(args.max_sentence_num)])
else:
sentences_lens.append([len(x) for x in tmp_words])
sentence_cut_words.append(tmp_words)
word_ids = [vocab.do_encode(x)[0] for x in tmp_words]
if args.padding_data:
if len(word_ids) < args.max_sentence_num:
for i in range(args.max_sentence_num - len(word_ids)):
word_ids.append([0])
word_ids = tf.keras.preprocessing.sequence.pad_sequences(word_ids,
maxlen=args.max_sequence_length,
padding="post",
truncating="post",
value=0)
assert np.max(word_ids) < args.vocab_size
assert np.max(agg_df["role"]) < 6
sentence_word_ids.append(word_ids)
return sentence_word_ids, roles, sentence_nums, sentences_lens
def _do_label_vectorize(df):
df = df.copy()
df.index = range(len(df))
df["sentence"] = df["sentence"].map(eval)
grouped = df.groupby("doc")
decoder_input_word_ids = []
decoder_output_word_ids = []
decoder_sentence_lens = []
for agg_name, agg_df in grouped:
question = {x for x in agg_df["question"]}
question_text = question.pop()
cut_words = eval(question_text)
decoder_input_word_ids.append(
vocab.do_encode(cut_words, mode="bos")[0]
)
decoder_output_word_ids.append(
vocab.do_encode(cut_words, mode="eos")[0]
)
decoder_sentence_lens.append(
len(cut_words) + 1
)
return decoder_input_word_ids, decoder_output_word_ids, decoder_sentence_lens
train_sentence_word_ids, train_roles, train_sentence_nums, train_sentences_lens = _do_vectorize(df_train, name=True)
dev_sentence_word_ids, dev_roles, dev_sentence_nums, dev_sentences_lens = _do_vectorize(df_dev, name=True)
test_sentence_word_ids, test_roles, test_sentence_nums, test_sentences_lens = _do_vectorize(df_test, name=True)
train_similar_word_ids, train_similar_roles, train_similar_nums, train_similar_lens = _do_vectorize(df_train_sim, name=True)
dev_similar_word_ids, dev_similar_roles, dev_similar_nums, dev_similar_lens = _do_vectorize(df_dev_sim, name=True)
test_similar_word_ids, test_similar_roles, test_similar_nums, test_similar_lens = _do_vectorize(df_test_sim, name=True)
train_decoder_input_word_ids, train_decoder_output_word_ids, train_decoder_sentence_lens = _do_label_vectorize(df_train)
dev_decoder_input_word_ids, dev_decoder_output_word_ids, dev_decoder_sentence_lens = _do_label_vectorize(df_dev)
test_decoder_input_word_ids, test_decoder_output_word_ids, test_decoder_sentence_lens = _do_label_vectorize(df_test)
with open(args.data_file, 'wb') as pkl_file:
data = [
list(zip(
train_sentence_word_ids, train_roles, train_sentence_nums, train_sentences_lens,
train_similar_word_ids, train_similar_roles, train_similar_nums, train_similar_lens,
train_decoder_input_word_ids, train_decoder_output_word_ids, train_decoder_sentence_lens)),
list(zip(
dev_sentence_word_ids, dev_roles, dev_sentence_nums, dev_sentences_lens,
dev_similar_word_ids, dev_similar_roles, dev_similar_nums, dev_similar_lens,
dev_decoder_input_word_ids, dev_decoder_output_word_ids, dev_decoder_sentence_lens)),
list(zip(
test_sentence_word_ids, test_roles, test_sentence_nums, test_sentences_lens,
test_similar_word_ids, test_similar_roles, test_similar_nums, test_similar_lens,
test_decoder_input_word_ids, test_decoder_output_word_ids, test_decoder_sentence_lens))
]
pickle.dump(data, pkl_file)
return data
if __name__ == '__main__':
args = model_opts()
load_data(args)