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
|
/*
* SPDX-FileCopyrightText: 2017-2017 CSSlayer <wengxt@gmail.com>
*
* SPDX-License-Identifier: LGPL-2.1-or-later
*/
#include "libime/core/constants.h"
#include "libime/core/datrie.h"
#include "libime/core/languagemodel.h"
#include <boost/algorithm/string.hpp>
#include <cmath>
#include <fcitx-utils/log.h>
#include <fstream>
#include <iostream>
#include <string>
#include <unistd.h>
#include <unordered_map>
#include <vector>
void usage(const char *argv0) {
std::cout
<< "Usage: " << argv0
<< " [-f <score>] [-s <maxSize>] <model.lm> <source.arpa> <dest>\n"
<< " -d <source.lm.predict> <dest>\n"
<< " -f: Set score filter\n"
<< " -s: Set max number of prediction per word\n"
<< " -h: Show this help\n";
}
float score(const libime::LanguageModel &model, std::string_view w,
std::string_view w2) {
return model.wordsScore(model.nullState(),
std::vector<std::string_view>{w, w2});
}
void parse(const char *modelFile, const char *arpa, const char *output,
float filter, unsigned long maxSize) {
using namespace libime;
LanguageModel model(modelFile);
DATrie<float> trie;
std::ifstream fin;
std::istream *in;
if (strcmp(arpa, "-") == 0) {
in = &std::cin;
} else {
fin.open(arpa, std::ios::in | std::ios::binary);
in = &fin;
}
std::string line;
std::unordered_map<std::string, std::unordered_map<std::string, float>>
word;
auto isSpaceCheck = boost::is_any_of(" \n\t\r\v\f");
int grams = -1;
while (!in->eof()) {
if (!std::getline(*in, line)) {
break;
}
boost::trim_if(line, isSpaceCheck);
if (line == "\\2-grams:") {
grams = 2;
continue;
}
if (line == "\\3-grams:") {
break;
}
if (grams < 0) {
continue;
}
std::vector<std::string> tokens;
boost::split(tokens, line, isSpaceCheck);
if (tokens.size() < static_cast<size_t>(grams) + 1) {
continue;
}
// We don't want prediction generate <unk>
if (std::find(tokens.begin() + 1, tokens.end(), "<unk>") !=
tokens.end()) {
continue;
}
auto s = score(model, tokens[grams - 1], tokens[grams]);
if (s > filter) {
word[tokens[grams - 1]][tokens[grams]] = s;
}
}
for (auto &p : word) {
std::vector<std::pair<std::string, float>> result(p.second.begin(),
p.second.end());
std::sort(result.begin(), result.end(),
[](const auto &lhs, const auto &rhs) {
if (lhs.second != rhs.second) {
return lhs.second > rhs.second;
}
return lhs.first < rhs.first;
});
if (result.size() > maxSize) {
result.resize(maxSize);
}
for (auto &p2 : result) {
trie.set(p.first + "|" + p2.first, p2.second);
}
}
FCITX_LOG(Info) << "Memory: " << trie.mem_size()
<< " Number of entries: " << trie.size();
std::ofstream fout;
std::ostream *out;
if (strcmp(output, "-") == 0) {
out = &std::cout;
} else {
fout.open(output, std::ios::out | std::ios::binary);
out = &fout;
}
trie.save(*out);
}
void dump(const char *input, const char *output) {
using namespace libime;
std::ifstream fin;
std::istream *in;
if (strcmp(input, "-") == 0) {
in = &std::cin;
} else {
fin.open(input, std::ios::in | std::ios::binary);
in = &fin;
}
DATrie<float> trie;
trie.load(*in);
std::ofstream fout;
std::ostream *out;
if (strcmp(output, "-") == 0) {
out = &std::cout;
} else {
fout.open(output, std::ios::out | std::ios::binary);
out = &fout;
}
trie.foreach([out, &trie](float value, size_t len, uint64_t pos) {
std::string s;
trie.suffix(s, len, pos);
*out << s << " " << value << std::endl;
return true;
});
}
int main(int argc, char *argv[]) {
int c;
unsigned long maxSize = 15;
float filter =
std::log10(libime::DEFAULT_LANGUAGE_MODEL_UNKNOWN_PROBABILITY_PENALTY) +
1;
bool doDump = false;
while ((c = getopt(argc, argv, "df:s:h")) != -1) {
switch (c) {
case 'd':
doDump = true;
break;
case 'f':
filter = std::stof(optarg);
break;
case 's':
maxSize = std::stoul(optarg);
break;
case 'h':
usage(argv[0]);
return 0;
default:
usage(argv[0]);
return 1;
}
}
if (doDump) {
if (optind + 1 >= argc) {
usage(argv[0]);
return 1;
}
dump(argv[optind], argv[optind + 1]);
} else {
if (optind + 2 >= argc) {
usage(argv[0]);
return 1;
}
parse(argv[optind], argv[optind + 1], argv[optind + 2], filter,
maxSize);
}
return 0;
}
|
