-
Notifications
You must be signed in to change notification settings - Fork 259
/
convert_language_model.py
executable file
·248 lines (219 loc) · 9.2 KB
/
convert_language_model.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
#!/usr/bin/env python
######################## BEGIN LICENSE BLOCK ########################
# Contributor(s):
# 10.02.2015 - helour - first attempt
# 08.08.2016 - Dan Blanchard - first usable release
#
# This library is free software; you can redistribute it and/or
# modify it under the terms of the GNU Lesser General Public
# License as published by the Free Software Foundation; either
# version 2.1 of the License, or (at your option) any later version.
#
# This library is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
# Lesser General Public License for more details.
#
# You should have received a copy of the GNU Lesser General Public
# License along with this library; if not, write to the Free Software
# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA
# 02110-1301 USA
#
######################### END LICENSE BLOCK #########################
"""
Convert old style SBCS model to new
"""
import os
import sys
from argparse import ArgumentDefaultsHelpFormatter, ArgumentParser
from string import ascii_letters
import chardet
from chardet import __version__
from chardet.metadata.languages import LANGUAGES
from chardet.sbcharsetprober import SingleByteCharSetModel
# Turn ascii_letters into a set to make other ops easier
ascii_letters = set(ascii_letters)
def normalize_name(charset_name):
"""Convert name to proper Python constant format"""
# Title case to start
charset_name = charset_name.upper()
# Underscores instead of hyphens
charset_name = charset_name.replace("-", "_")
return charset_name
def convert_sbcs_model(old_model, alphabet):
"""Create a SingleByteCharSetModel object representing the charset."""
# Setup tables necessary for computing transition frequencies for model
char_to_order = {i: order for i, order in enumerate(old_model["char_to_order_map"])}
pos_ratio = old_model["typical_positive_ratio"]
keep_ascii_letters = old_model["keep_english_letter"]
curr_model = SingleByteCharSetModel(
charset_name=old_model["charset_name"],
language=old_model["language"],
char_to_order_map=char_to_order,
# language_model is filled in later
language_model=None,
typical_positive_ratio=pos_ratio,
keep_ascii_letters=keep_ascii_letters,
alphabet=alphabet,
)
return curr_model
def print_char_to_order(var_name, order_map, charset_name, output_file):
print(f"{var_name} = {{", file=output_file)
for char, order in sorted(order_map.items()):
char_bytes = bytes(bytearray((char,)))
try:
unicode_char = char_bytes.decode(charset_name)
except UnicodeError:
unicode_char = None
print(f" {char!r}: {order!r}, # {unicode_char!r}", file=output_file)
print("}\n", file=output_file)
def print_language_model(var_name, language_model, output_file, char_ranks):
print(
"# 3: Positive\n" "# 2: Likely\n" "# 1: Unlikely\n" "# 0: Negative\n",
file=output_file,
)
print(f"{var_name} = {{", file=output_file)
for first_char, sub_dict in sorted(language_model.items()):
# Skip empty sub_dicts
if not sub_dict or first_char not in char_ranks:
continue
print(f" {char_ranks[first_char]!r}: {{ # {first_char!r}", file=output_file)
for second_char, likelihood in sorted(sub_dict.items()):
if second_char not in char_ranks:
continue
print(
f" {char_ranks[second_char]!r}: {likelihood!r}, # "
f"{second_char!r}",
file=output_file,
)
print(" },", file=output_file)
print("}\n", file=output_file)
def convert_models_for_lang(language):
"""Convert old SingleByteCharSetModels for the given language"""
# Validate language
language = language.title()
lang_metadata = LANGUAGES.get(language)
if not lang_metadata:
raise ValueError(
f"Unknown language: {language}. If you are adding a model for a"
" new language, you must first update metadata/"
"languages.py"
)
lang_mod_name = f"lang{language.lower()}model"
if not os.path.exists(os.path.join("chardet", lang_mod_name + ".py")):
print(f"Skipping {language} because it does not have an old model.")
return
lang_mod = getattr(chardet, lang_mod_name)
print(
f"\n{language}\n----------------------------------------------------------------"
)
print(f"Keep ASCII Letters: {lang_metadata.use_ascii}")
print(f"Alphabet: {lang_metadata.alphabet}")
# Create char-to-order maps (aka char-to-rank dicts)
charset_models = {}
char_ranks = {}
order_to_chars = {}
for var_name in dir(lang_mod):
if not ("Model" in var_name and "LangModel" not in var_name):
continue
old_model = getattr(lang_mod, var_name)
charset_name = old_model["charset_name"]
print(f"Converting charset model for {charset_name}")
sys.stdout.flush()
charset_models[charset_name] = convert_sbcs_model(
old_model, lang_metadata.alphabet
)
# Since we don't know which charsets have which characters, we have to
# try to reconstruct char_ranks (for letters only, since that's all
# the old language models contain)
for byte_hex, order in charset_models[charset_name].char_to_order_map.items():
# order 64 was basically ignored before because of the off by one
# error, but it's hard to know if training took that into account
if order > 64:
continue
# Convert to bytes in Python 2 and 3
char = bytes(bytearray((byte_hex,)))
try:
unicode_char = char.decode(charset_name)
except UnicodeDecodeError:
continue
if unicode_char not in char_ranks:
char_ranks[unicode_char] = order
order_to_chars[order] = unicode_char
elif char_ranks[unicode_char] != order:
raise ValueError(f"Unstable character ranking for {unicode_char}")
old_lang_model = getattr(lang_mod, f"{language.title()}LangModel")
language_model = {}
# Preserve off-by-one error here by ignoring first column and row
for i in range(1, 64):
if i not in order_to_chars:
continue
lang_char = order_to_chars[i]
language_model[lang_char] = {}
for j in range(1, 64):
if j not in order_to_chars:
continue
lang_char2 = order_to_chars[j]
language_model[lang_char][lang_char2] = old_lang_model[(i * 64) + j]
# Write output files
print(f"Writing output file for {language}\n\n")
sys.stdout.flush()
with open(f"lang{language.lower()}model.py", "w") as output_file:
upper_lang = language.upper()
# print header to set encoding
print(
"#!/usr/bin/env python\n"
"# -*- coding: utf-8 -*-\n\n"
"from chardet.sbcharsetprober import SingleByteCharSetModel\n\n",
file=output_file,
)
lm_name = f"{upper_lang}_LANG_MODEL"
print_language_model(lm_name, language_model, output_file, char_ranks)
print(
"# 255: Undefined characters that did not exist in training text\n"
"# 254: Carriage/Return\n"
"# 253: symbol (punctuation) that does not belong to word\n"
"# 252: 0 - 9\n"
"# 251: Control characters\n\n"
"# Character Mapping Table(s):",
file=output_file,
)
for charset_name, sbcs_model in charset_models.items():
normal_name = normalize_name(charset_name)
char_to_order_name = f"{normal_name}_{upper_lang}_CHAR_TO_ORDER"
print_char_to_order(
char_to_order_name,
sbcs_model.char_to_order_map,
charset_name,
output_file,
)
sbcs_model_name = f"{normal_name}_{upper_lang}_MODEL"
sbcs_model.char_to_order_map.clear()
sbcs_model_repr = (
repr(sbcs_model)
.replace("None", lm_name)
.replace("{}", char_to_order_name)
.replace(", ", (",\n" + " " * (len(sbcs_model_name) + 26)))
)
print(f"{sbcs_model_name} = {sbcs_model_repr}\n", file=output_file)
def main():
parser = ArgumentParser(
description=__doc__, formatter_class=ArgumentDefaultsHelpFormatter
)
parser.add_argument(
"language",
help="The name of the language the input documents are "
"in. Also the name of the language the generated "
"model will detect. If no language is specified, "
"models for all languages known to chardet will be"
" trained.",
nargs="*",
)
parser.add_argument("--version", action="version", version=__version__)
args = parser.parse_args()
if not args.language:
args.language = list(sorted(LANGUAGES.keys()))
for language in args.language:
convert_models_for_lang(language)
if __name__ == "__main__":
main()