This repository has been archived by the owner on Oct 31, 2023. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 143
/
Copy pathtext_normalizer.py
189 lines (166 loc) · 4.81 KB
/
text_normalizer.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
# Copyright (c) Facebook, Inc. and its affiliates.
#
# This source code is licensed under the MIT license found in the
# LICENSE file in the root directory of this source tree.
#
import re
import unicodedata
UNICODE_PUNCT = {
"，": ",",
"。": ".",
"、": ",",
"„": '"',
"”": '"',
"“": '"',
"«": '"',
"»": '"',
"１": '"',
"」": '"',
"「": '"',
"《": '"',
"》": '"',
"´": "'",
"∶": ":",
"：": ":",
"？": "?",
"！": "!",
"（": "(",
"）": ")",
"；": ";",
"–": "-",
"—": " - ",
"．": ". ",
"～": "~",
"’": "'",
"…": "...",
"━": "-",
"〈": "<",
"〉": ">",
"【": "[",
"】": "]",
"％": "%",
"►": "-",
}
UNICODE_PUNCT_RE = re.compile(f"[{''.join(UNICODE_PUNCT.keys())}]")
def replace_unicode_punct(text: str) -> str:
return "".join((UNICODE_PUNCT.get(c, c) for c in text))
def remove_unicode_punct(text: str) -> str:
"""More aggressive version of replace_unicode_punct but also faster."""
return UNICODE_PUNCT_RE.sub("", text)
def strip_accents(line: str) -> str:
"""Strips accents from a piece of text."""
nfd = unicodedata.normalize("NFD", line)
output = [c for c in nfd if unicodedata.category(c) != "Mn"]
if len(output) == line:
return line
return "".join(output)
# Build a regex matching all control characters.
NON_PRINTING_CHARS_RE = re.compile(
f"[{''.join(map(chr, list(range(0,32)) + list(range(127,160))))}]"
)
DIGIT_RE = re.compile(r"\d")
PUNCT_OR_NON_PRINTING_CHARS_RE = re.compile(
(UNICODE_PUNCT_RE.pattern + NON_PRINTING_CHARS_RE.pattern).replace("][", "")
)
def remove_non_printing_char(text: str) -> str:
return NON_PRINTING_CHARS_RE.sub("", text)
def normalize_spacing_for_tok(text: str, language: str = "en") -> str:
res = (
text.replace("\r", "")
# remove extra spaces
.replace("(", " (")
.replace(")", ") ")
.replace(" +", " ")
)
res = re.sub(r"\) ([\.\!\:\?\;\,])", r"\)\1", res)
res = res.replace("( ", "(").replace(" )", ")")
res = re.sub(r"(\d) \%", r"\1\%", res)
res = res.replace(" :", ":").replace(" ;", ";")
res = res.replace("`", "'").replace("''", ' " ')
res = (
res.replace("„", '"')
.replace("“", '"')
.replace("”", '"')
.replace("–", "-")
.replace("—", " - ")
.replace(" +", " ")
.replace("´", "'")
.replace("([a-z])‘([a-z])", r"\1'\2/")
.replace("([a-z])’([a-z])", r"\1'\2/")
.replace("‘", '"')
.replace("‚", '"')
.replace("’", '"')
.replace("''", '"')
.replace("´´", '"')
.replace("…", "...")
# French quotes
.replace(" « ", ' "')
.replace("« ", '"')
.replace("«", '"')
.replace(" » ", '" ')
.replace(" »", '"')
.replace("»", '"')
# handle pseudo-spaces
.replace(" %", "%")
.replace("nº ", "nº ")
.replace(" :", ":")
.replace(" ºC", " ºC")
.replace(" cm", " cm")
.replace(" ?", "?")
.replace(" !", "!")
.replace(" ;", ";")
.replace(", ", ", ")
.replace(" +", " ")
.replace("．", ". ")
)
# English "quotation," followed by comma, style
if language == "en":
res = re.sub(r"\"([,\.]+)", r"\1\"", res)
# Czech is confused
elif language == "cs" or language == "cz":
pass
# German/Spanish/French "quotation", followed by comma, style
else:
res = res.replace(',"', '",')
res = re.sub(
r"(\.+)\"(\s*[^<])", r"\"\1\2", res
) # don't fix period at end of sentence
if (
language == "de"
or language == "es"
or language == "cz"
or language == "cs"
or language == "fr"
):
res = re.sub(r"(\d) (\d)", r"\1,\2", res)
else:
res = re.sub(r"(\d) (\d)", r"\1.\2", res)
return res
def normalize(line: str, accent=True, case=True, numbers=True, punct=1) -> str:
line = line.strip()
if not line:
return line
if case:
line = line.lower()
if accent:
line = strip_accents(line)
if numbers:
line = DIGIT_RE.sub("0", line)
if punct == 1:
line = replace_unicode_punct(line)
elif punct == 2:
line = remove_unicode_punct(line)
line = remove_non_printing_char(line)
return line
def slow_normalize_for_dedup(line: str) -> str:
return normalize(line, accent=False, case=True, numbers=True, punct=2)
def normalize_for_dedup(line: str) -> str:
line = line.strip()
if not line:
return line
# case
line = line.lower()
# numbers
line = DIGIT_RE.sub("0", line)
line = PUNCT_OR_NON_PRINTING_CHARS_RE.sub("", line)
return line