-
Notifications
You must be signed in to change notification settings - Fork 3
/
patch_classifier.py
263 lines (245 loc) · 10.5 KB
/
patch_classifier.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
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
# coding=utf-8
# @author: cer
# 有一些分类器分错的明显的错误，使用规则矫正。
# 这个脚本跟在分类器脚本后面跑。
from __future__ import print_function
import pandas as pd
import numpy as np
import re
import pickle as pkl
import time
pred_file = "output/class_pred_16.v2.5.csv"
out_file = "output/class_pred_16_fixed.v2.5.csv"
all_pred_file_name = "output/train_pred.v2.5.csv"
all_output_file_name = "output/train_pred.v2.5.fixed.2.csv"
change_file = "output/patch_change.csv"
train_file = "input/en_train.csv"
test_file = "input/en_test_2.csv"
labels = ['PLAIN', 'PUNCT', 'DATE', 'LETTERS', 'CARDINAL', 'VERBATIM',
'DECIMAL', 'MEASURE', 'MONEY', 'ORDINAL', 'TIME', 'ELECTRONIC',
'DIGIT', 'FRACTION', 'TELEPHONE', 'ADDRESS']
label2index = dict(zip(labels, range(len(labels))))
index2label = dict([(v, k) for k, v in label2index.items()])
# with open("output/letter_set.pkl", "rb") as f:
# letter_set = pkl.load(f)
letters_name = "output/letter_dict.pkl"
with open(letters_name, "rb") as f:
letter_dict = pkl.load(f)
# chem_dict = {"CO₂": "carbon di oxide",
# "ZnI₂": "zinc iodide",
# "NH₄": "ammonium",
# "SiO₂": "silicon di oxide",
# "CaCO₃": "calcium carbonate",
# "CaCO3": "calcium carbonate",
# "H₂O": "water",
# "TiO2": "rutile"}
chem_dict = {"CO₂": "CO₂",
"ZnI₂": "ZnI₂",
"NH₄": "NH₄",
"SiO₂": "SiO₂",
"OH)₂": "OH)₂",
"CaCO₃": "CaCO₃",
"CaCO3": "CaCO3",
"H₂O": "H₂O",
"NH₃": "NH₃",
"C₂₅H₂₁": "C₂₅H₂₁",
"XeF₂": "XeF₂",
"N₅O": "N₅O",
"CH₃": "CH₃",
"PI₃": "PI₃",
"Cu₃O₇": "Cu₃O₇",
"CH₂": "CH₂",
"Zn(OH)₂": "Zn(OH)₂",
"NH₂": "NH₂",
"MnS₂": "MnS₂",
"C₆H₂": "C₆H₂",
"C₁₉H₁₆": "C₁₉H₁₆",
"AlCl₃": "AlCl₃",
"SO₂": "SO₂",
"CCl₄": "CCl₄",
"C₁₀H₆": "C₁₀H₆",
"Ca₃(AsO₄)₂": "Ca₃(AsO₄)₂",
"BaTiO3": "BaTiO3",
"F₁F₂": "F₁F₂",
"C₃H": "C₃H",
"SiO4": "SiO4",
"CO)₃": "CO)₃",
"O₂N)₃": "O₂N)₃",
"SiO₅": "SiO₅",
"Ca5(PO4)3": "Ca5(PO4)3",
"Na(SO4)2": "Na(SO4)2",
"Mg₃B₇": "Mg₃B₇",
"O₁₃Cl": "O₁₃Cl",
"Li2O2": "Li2O2",
"Li2O": "Li2O",
"HNO₃": "HNO₃",
"AlSi₃": "AlSi₃",
"TiO2": "TiO2",
"BeF₂": "BeF₂",
"PbCl₂": "PbCl₂",
"C₄H₄": "C₄H₄",
"CO)₆": "CO)₆",
"TiH1": "TiH1",
"Pb₃As₄": "Pb₃As₄",
"C₆H₄": "C₆H₄",
}
dash_pat = re.compile(r"^[a-zA-Z]+-[a-zA-Z]+$")
def is_com(token):
token = str(token)
pt = re.compile(r"\.[a-zA-Z]{2,}")
return True if pt.search(token) else False
def is_year(token):
token = str(token)
return token.isdigit() and len(token) == 4 and 1000 <= int(token) <= 2099
def has_measure_string(token):
token = str(token)
mea_strs = ['years', 'year', 'months', 'month', 'weeks', 'week', 'days', 'day',
'hours', 'minutes', 'min', 'seconds', 'nautical', 'acres', 'hectares',
'tonnes', 'tons', 'kilobytes', 'nanobarns', 'square', 'miles', 'kilometres',
'kilometre', 'kilometers', 'kilometer', 'meters', 'meter', 'metres', 'metre',
'feet', 'yards', 'barrels', 'calories', 'watts', 'knots', 'inches', 'pounds',
'degrees', 'bar', 'Cellos']
for one in mea_strs:
if one in token:
return True
return False
def ends_with_measure_string(token):
token = str(token)
mea_strs = ['years', 'year', 'months', 'month', 'weeks', 'week', 'days',
'hours', 'minutes', 'min', 'seconds', 'nautical', 'acres', 'hectares',
'tonnes', 'tons', 'kilobytes', 'nanobarns', 'square', 'miles', 'kilometres',
'kilometre', 'kilometers', 'kilometer', 'meters', 'meter', 'metres', 'metre',
'feet', 'yards', 'barrels', 'calories', 'watts', 'knots', 'inches', 'pounds',
'degrees', 'bar', 'Cellos']
for one in mea_strs:
if token.endswith(one):
return True
return False
def has_dash(token):
token = str(token)
if dash_pat.search(token):
return True
else:
return False
def patch():
test_df = pd.read_csv(test_file)
pred_df = pd.read_csv(pred_file)
pred_df["before"] = test_df["before"]
print("start iterating...")
data_num = pred_df.shape[0]
count = 0
change_list = []
s = time.time()
for i, row in pred_df.iterrows():
# 1.被误认为是cardinal的实际上的date的年份
if index2label[row["class_pred"]] != "DATE" and is_year(row["before"]):
# 不能直接修改row，因为row是副本
pred_df.loc[i, "class_pred"] = label2index["DATE"]
row["class_before"] = index2label[row["class_pred"]]
row["class_pred"] = "DATE"
change_list.append(row)
elif index2label[row["class_pred"]] == "PLAIN":
# 2.被误认为是plain的实际上的electronic的网站
if is_com(row["before"]):
pred_df.loc[i, "class_pred"] = label2index["ELECTRONIC"]
row["class_pred"] = "ELECTRONIC"
row["class_before"] = "PLAIN"
change_list.append(row)
# 3.被误认为是plain的实际上是letters
elif row["before"] in letter_dict["LETTERS"]:
if row["before"] not in letter_dict["PLAIN"]:
change = True
elif letter_dict["LETTERS"][row["before"]] > letter_dict["PLAIN"][row["before"]]:
change = True
else:
change = False
if change:
pred_df.loc[i, "class_pred"] = label2index["LETTERS"]
row["class_pred"] = "LETTERS"
row["class_before"] = "PLAIN"
change_list.append(row)
# 3.被误认为是letters的实际上是plain的
elif index2label[row["class_pred"]] == "LETTERS":
# if row["before"] not in letter_dict["LETTERS"] or letter_dict["LETTERS"][row["before"]] \
# < letter_dict["PLAIN"][row["before"]]:
if row["before"] in chem_dict:
pred_df.loc[i, "class_pred"] = label2index["PLAIN"]
row["class_pred"] = "PLAIN"
row["class_before"] = "LETTERS"
change_list.append(row)
# 4.被误认为是decimal的实际上是measured
elif index2label[row["class_pred"]] == "DECIMAL":
if has_measure_string(row["before"]):
pred_df.loc[i, "class_pred"] = label2index["MEASURE"]
row["class_pred"] = "MEASURE"
row["class_before"] = "DECIMAL"
change_list.append(row)
# 5.被误认为是date的实际上是measured
elif index2label[row["class_pred"]] == "DATE":
if ends_with_measure_string(row["before"]):
pred_df.loc[i, "class_pred"] = label2index["MEASURE"]
row["class_pred"] = "MEASURE"
row["class_before"] = "DATE"
change_list.append(row)
# 6.被误认为是money的实际上是measured
elif index2label[row["class_pred"]] == "MONEY":
if has_measure_string(row["before"]):
pred_df.loc[i, "class_pred"] = label2index["MEASURE"]
row["class_pred"] = "MEASURE"
row["class_before"] = "MONEY"
change_list.append(row)
# 7.被误认为是cardinal的实际上是measured
elif index2label[row["class_pred"]] == "CARDINAL":
if ends_with_measure_string(row["before"]):
pred_df.loc[i, "class_pred"] = label2index["MEASURE"]
row["class_pred"] = "MEASURE"
row["class_before"] = "CARDINAL"
change_list.append(row)
# 8.被误认为是electronic的实际上是plain
elif index2label[row["class_pred"]] == "ELECTRONIC":
if has_dash(row["before"]):
pred_df.loc[i, "class_pred"] = label2index["PLAIN"]
row["class_pred"] = "PLAIN"
row["class_before"] = "ELECTRONIC"
change_list.append(row)
print("time costs: ", time.time() - s)
print("changed row num: ", len(change_list))
change_df = pd.DataFrame(change_list)
change_df.to_csv(change_file, index=False)
print("save change rows to file: ", change_file)
pred_df.to_csv(out_file, index=False)
print("save fixed prediction to file: ", out_file)
print("done!")
def patch_train_file():
pred_df = pd.read_csv(all_pred_file_name)
print("start iterating...")
data_num = pred_df.shape[0]
count = 0
for i, row in pred_df.iterrows():
# 1.被误认为是cardinal的实际上的date的年份
if row["class_pred"] == "CARDINAL":
if is_year(row["before"]):
# if is_year(row["before"]) and i + 1 < data_num and pred_df.loc[i + 1, "before"] not in ["-", "/"]:
# 不能直接修改row，因为row是副本
pred_df.loc[i, "class_pred"] = "DATE"
count += 1
elif row["class_pred"] == "PLAIN":
# 2.被误认为是plain的实际上的electronic的网站
if is_com(row["before"]):
pred_df.loc[i, "class_pred"] = "ELECTRONIC"
count += 1
# 3.被误认为是plain的实际上是letters
elif row["before"] in letter_set:
pred_df.loc[i, "class_pred"] = "LETTERS"
count += 1
elif row["class_pred"] == "LETTERS":
if row["before"] not in letter_set:
pred_df.loc[i, "class_pred"] = "PLAIN"
count += 1
print("changed row num: ", count)
pred_df.to_csv(all_output_file_name, index=False)
print("save fixed prediction to file: ", all_output_file_name)
print("done!")
if __name__ == '__main__':
patch()
# patch_train_file()