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
|
#include "Scorer.h"
#include <limits>
#include "Vocabulary.h"
#include "Util.h"
#include "Singleton.h"
#include "util/tokenize_piece.hh"
#if defined(__GLIBCXX__) || defined(__GLIBCPP__)
#include "PreProcessFilter.h"
#endif
using namespace std;
namespace MosesTuning
{
namespace
{
// For tokenizing a hypothesis translation, we may encounter unknown tokens which
// do not exist in the corresponding reference translations.
const int kUnknownToken = -1;
} // namespace
Scorer::Scorer(const string& name, const string& config)
: m_name(name),
m_vocab(mert::VocabularyFactory::GetVocabulary()),
#if defined(__GLIBCXX__) || defined(__GLIBCPP__)
m_filter(NULL),
#endif
m_score_data(NULL),
m_enable_preserve_case(true)
{
InitConfig(config);
}
Scorer::~Scorer()
{
Singleton<mert::Vocabulary>::Delete();
#if defined(__GLIBCXX__) || defined(__GLIBCPP__)
delete m_filter;
#endif
}
void Scorer::InitConfig(const string& config)
{
// cerr << "Scorer config string: " << config << endl;
size_t start = 0;
while (start < config.size()) {
size_t end = config.find(",", start);
if (end == string::npos) {
end = config.size();
}
string nv = config.substr(start, end - start);
size_t split = nv.find(":");
if (split == string::npos) {
throw runtime_error("Missing colon when processing scorer config: " + config);
}
const string name = nv.substr(0, split);
const string value = nv.substr(split + 1, nv.size() - split - 1);
cerr << "name: " << name << " value: " << value << endl;
m_config[name] = value;
start = end + 1;
}
}
void Scorer::TokenizeAndEncode(const string& line, vector<int>& encoded) const
{
for (util::TokenIter<util::AnyCharacter, true> it(line, util::AnyCharacter(" "));
it; ++it) {
if (!m_enable_preserve_case) {
string token = it->as_string();
for (std::string::iterator sit = token.begin();
sit != token.end(); ++sit) {
*sit = tolower(*sit);
}
encoded.push_back(m_vocab->Encode(token));
} else {
encoded.push_back(m_vocab->Encode(it->as_string()));
}
}
}
void Scorer::TokenizeAndEncodeTesting(const string& line, vector<int>& encoded) const
{
for (util::TokenIter<util::AnyCharacter, true> it(line, util::AnyCharacter(" "));
it; ++it) {
if (!m_enable_preserve_case) {
string token = it->as_string();
for (std::string::iterator sit = token.begin();
sit != token.end(); ++sit) {
*sit = tolower(*sit);
}
mert::Vocabulary::const_iterator cit = m_vocab->find(token);
if (cit == m_vocab->end()) {
encoded.push_back(kUnknownToken);
} else {
encoded.push_back(cit->second);
}
} else {
mert::Vocabulary::const_iterator cit = m_vocab->find(it->as_string());
if (cit == m_vocab->end()) {
encoded.push_back(kUnknownToken);
} else {
encoded.push_back(cit->second);
}
}
}
}
/**
* Set the factors, which should be used for this metric
*/
void Scorer::setFactors(const string& factors)
{
if (factors.empty()) return;
vector<string> factors_vec;
split(factors, '|', factors_vec);
for(vector<string>::iterator it = factors_vec.begin(); it != factors_vec.end(); ++it) {
int factor = atoi(it->c_str());
m_factors.push_back(factor);
}
}
/**
* Set unix filter, which will be used to preprocess the sentences
*/
void Scorer::setFilter(const string& filterCommand)
{
if (filterCommand.empty()) return;
#if defined(__GLIBCXX__) || defined(__GLIBCPP__)
m_filter = new PreProcessFilter(filterCommand);
#else
throw runtime_error("Cannot use filter command as mert was compiled with non-gcc compiler");
#endif
}
/**
* Take the factored sentence and return the desired factors
*/
string Scorer::applyFactors(const string& sentence) const
{
if (m_factors.size() == 0) return sentence;
vector<string> tokens;
split(sentence, ' ', tokens);
stringstream sstream;
for (size_t i = 0; i < tokens.size(); ++i) {
if (tokens[i] == "") continue;
vector<string> factors;
split(tokens[i], '|', factors);
int fsize = factors.size();
if (i > 0) sstream << " ";
for (size_t j = 0; j < m_factors.size(); ++j) {
int findex = m_factors[j];
if (findex < 0 || findex >= fsize) throw runtime_error("Factor index is out of range.");
if (j > 0) sstream << "|";
sstream << factors[findex];
}
}
return sstream.str();
}
/**
* Preprocess the sentence with the filter (if given)
*/
string Scorer::applyFilter(const string& sentence) const
{
#if defined(__GLIBCXX__) || defined(__GLIBCPP__)
if (m_filter) {
return m_filter->ProcessSentence(sentence);
} else {
return sentence;
}
#endif
return sentence;
}
float Scorer::score(const candidates_t& candidates) const
{
diffs_t diffs;
statscores_t scores;
score(candidates, diffs, scores);
return scores[0];
}
}
|
