-
Notifications
You must be signed in to change notification settings - Fork 9
/
filter_line_sample1.py
198 lines (180 loc) · 7.42 KB
/
filter_line_sample1.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
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
import os
import re
import sys
import string
import random
import textwrap
import unicodedata
LEFT_PUNCTS = "\"'(([{‘“〈《「『【〔"
RIGHT_PUNCTS = "\"'))]}’”〉》」』】〕"
MIDDLE_PUNCTS = "!&*,./:;?@\\_~·、。々!,:;?—…"
DIGIT_PUNCTS = ("#$¥", "+-*/×÷=≠≈<>≤≥^", "%°")
ALL_PUNCTS = (LEFT_PUNCTS + RIGHT_PUNCTS + MIDDLE_PUNCTS + ''.join(DIGIT_PUNCTS))
re_noleft = re.compile(r'^[!%),.:;>?\]}°·’”、。〉》」』】〕!),:;?]+')
re_noright = re.compile(r'[$\(\[{¥·‘“〈《「『【〔(]+$')
re_empty_pair = re.compile(r'[%s]+[%s]*[%s]+' %
tuple(map(re.escape, (LEFT_PUNCTS[2:], MIDDLE_PUNCTS, RIGHT_PUNCTS[2:]))))
re_qs = re.compile('\?\s*\?|?\s*?|[\ufffc\ufffd]')
re_open = re.compile(r'([%s])\s*[%s]+' % tuple(map(re.escape,
(LEFT_PUNCTS[2:], MIDDLE_PUNCTS + DIGIT_PUNCTS[2]))))
re_close = re.compile(r'[&*,/:;@\\_·、,:;%s]+\s*([%s])' % tuple(map(re.escape,
(DIGIT_PUNCTS[0] + DIGIT_PUNCTS[1], RIGHT_PUNCTS[2:]))))
re_dup = re.compile(r'[!,:;?@\\·、。!,:;? ]{2,}')
re_exqm = re.compile(r'^[!?]{1,5}\s*$|^[!?]{1,5}$')
re_space = re.compile(r'\s+')
re_fwspace_left = re.compile(r'([(‘“〈《「『【〔]+)\s+')
re_fwspace_right = re.compile(r'\s+([)’”〉》」』】〕]+)')
re_fwspace_mid = re.compile(r'\s+([·、。々!,:;?—…]+)\s+')
re_allpuncts = re.compile(r'[%s ]+' % re.escape(''.join(sorted(set(ALL_PUNCTS)))))
RE_WS_IN_FW = re.compile(
r'([‘’“”…─\u2e80-\u312f\u3200-\u32ff\u3400-\u4dbf\u4e00-\u9fff'
'\uf900-\ufaff\ufe30-\ufe57\uff00-\uffef\U00020000-\U0002A6DF'
'\U0002A700-\U0002EBEF\U00030000-\U0003134F'
'\U0002F800-\U0002FA1F])\s+(?=[‘’“”…\u2e80-\u312f\u3200-\u32ff'
'\u3400-\u4dbf\u4e00-\u9fff\uf900-\ufaff\ufe30-\ufe57\uff00-\uffef'
'\U00020000-\U0002A6DF\U0002A700-\U0002EBEF'
'\U00030000-\U0003134F\U0002F800-\U0002FA1F])')
detokenize = lambda s: RE_WS_IN_FW.sub(r'\1', s)
random.seed(12345)
class CJKTextWrapper(textwrap.TextWrapper):
wordsep_simple_re = re.compile(r'(\s+|[%s]*[\u4e00-\u9FFF][%s]*|[¥$+-]*\d+(?:\.\d+)?(?:e[+-]\d+)?)' % (
re.escape(LEFT_PUNCTS), re.escape(RIGHT_PUNCTS + MIDDLE_PUNCTS.replace('¥', ''))
))
def _split(self, text):
chunks = self.wordsep_simple_re.split(text)
return [c for c in chunks if c]
def remove_dup_puncts(match):
s = match.group(0)
if s.rstrip() == s[0] and s[0] in '!,:;?':
return s
elif re_exqm.match(s):
return s
elif not s.strip():
return ' '
return s.strip()[0]
def remove_fw_space(x):
if ' ' not in x:
return x
merged = []
last_width = None
for frag in x.split():
this_width = (unicodedata.east_asian_width(frag[0]) in 'FW')
if last_width is None or this_width == last_width == True:
merged.append(frag)
else:
merged.append(' ' + frag)
last_width = (unicodedata.east_asian_width(frag[-1]) in 'FW')
return ''.join(merged)
def load_simplewordlist(filename):
words = []
with open(filename, 'r', encoding='utf-8') as f:
for ln in f:
words.append(ln.strip())
return words
if __name__ == '__main__':
script_dir = os.path.abspath(os.path.dirname(__file__))
char_set = set(ALL_PUNCTS)
char_set.add(' ')
unusual_set = set(char_set)
char_set.update(string.ascii_letters)
char_set.update(string.digits)
if len(sys.argv) < 2 or sys.argv[1] == 'zhs':
lang = 'chi_sim'
filename_chlist = 'wordlist/zhs_chars.txt'
filename_phonetic = 'wordlist/pinyin.txt'
else:
lang = 'chi_tra'
filename_chlist = 'wordlist/zht_chars.txt'
filename_phonetic = 'wordlist/bopomofo.txt'
chlist = load_simplewordlist(
os.path.join(script_dir, filename_chlist))
char_set.update(chlist)
unusual_set.update(chlist[4000:])
char_set.add('〇')
phlist = ''.join(load_simplewordlist(
os.path.join(script_dir, filename_phonetic)))
unusual_set.update(set(phlist).difference(string.ascii_letters))
char_set.update(phlist)
max_length = 50
unseen_lengths = {ch: set(range(1, max_length+1)) for ch in char_set}
seen_lines = set()
for raw_ln in sys.stdin.buffer:
try:
long_ln = raw_ln.strip().decode('utf-8')
except UnicodeError:
continue
if '??' in long_ln:
continue
long_ln = detokenize(re_space.sub(' ', long_ln))
long_ln = re_fwspace_mid.sub(r'\1', long_ln)
long_ln = re_fwspace_left.sub(r'\1', long_ln)
long_ln = re_fwspace_right.sub(r'\1', long_ln)
long_ln = re_empty_pair.sub('', long_ln)
long_ln = re_open.sub(r'\1', long_ln)
long_ln = re_close.sub(r'\1', long_ln)
long_ln = long_ln.replace('...', '…').replace('… …', '……')
long_ln = re_dup.sub(remove_dup_puncts, long_ln)
long_ln = remove_fw_space(long_ln)
if ' ' in long_ln:
merged = []
last_width = None
for frag in long_ln.split():
this_width = unicodedata.east_asian_width(frag[0])
if last_width is None or this_width == last_width == True:
merged.append(frag)
else:
merged.append(' ' + frag)
last_width = this_width
if len(long_ln) < max_length:
lns = (long_ln,)
else:
wrapper = CJKTextWrapper(width=round(random.triangular(30, max_length, max_length)))
lns = wrapper.wrap(long_ln)
for ln in lns:
if not ln:
continue
elif re_qs.search(ln):
continue
result = []
unknown = 0
unusual = 0
remove_chars = set()
for ch in ln:
if ch not in char_set:
unknown += 1
category = unicodedata.category(ch)
if category[0] == 'Lo':
result.append(ch)
elif ch == '\u3000' or category[0] in 'PS':
result.append(' ')
continue
unusual += (ch in unusual_set)
result.append(ch)
remove_chars.add(ch)
result_line = ''.join(result).strip()
result_line = re_noright.sub('', re_noleft.sub('', result_line))
if not result_line:
continue
result_length = len(result_line)
if ((len(set(result_line)) == 1 and result_length > 2) or
unknown / result_length >= 0.5 or
(unusual / result_length > 0.5 and result_length > 2)
):
continue
has_unseen = False
for ch in remove_chars:
lengths = unseen_lengths.get(ch, ())
if result_length in lengths:
lengths.remove(result_length)
has_unseen = True
unique_ratio = len(set(result_line)) / result_length
if (has_unseen or
# or random.random() < 0.0015):
result_length >= 40 and random.random() < 0.005 * unique_ratio):
result_charonly = re_allpuncts.sub('', result_line)
if result_charonly in seen_lines:
continue
seen_lines.add(result_charonly)
print(result_line)