blob: 6314190234ea16644e8a3e0863ea54169b8e0d69 (
plain)
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
|
// $Id$
#include <iostream>
#include <sstream>
#include <algorithm>
#include "moses/StaticData.h"
#include "moses/Hypothesis.h"
#include "moses/ChartHypothesis.h"
#include "TranslationAnalysis.h"
#include "moses/FF/StatefulFeatureFunction.h"
#include "moses/FF/StatelessFeatureFunction.h"
#include "moses/LM/Base.h"
#include "util/string_stream.hh"
using namespace Moses;
namespace TranslationAnalysis
{
void PrintTranslationAnalysis(std::ostream &os, const Hypothesis* hypo)
{
os << std::endl << "TRANSLATION HYPOTHESIS DETAILS:" << std::endl;
std::vector<const Hypothesis*> translationPath;
while (hypo) {
translationPath.push_back(hypo);
hypo = hypo->GetPrevHypo();
}
std::reverse(translationPath.begin(), translationPath.end());
std::vector<std::string> droppedWords;
std::vector<const Hypothesis*>::iterator tpi = translationPath.begin();
if(tpi == translationPath.end())
return;
++tpi; // skip initial translation state
std::vector<std::string> sourceMap;
std::vector<std::string> targetMap;
std::vector<unsigned int> lmAcc(0);
size_t lmCalls = 0;
bool doLMStats = ((*tpi)->GetLMStats() != 0);
if (doLMStats)
lmAcc.resize((*tpi)->GetLMStats()->size(), 0);
for (; tpi != translationPath.end(); ++tpi) {
util::StringStream sms;
util::StringStream tms;
std::string target = (*tpi)->GetTargetPhraseStringRep();
std::string source = (*tpi)->GetSourcePhraseStringRep();
Range twr = (*tpi)->GetCurrTargetWordsRange();
Range swr = (*tpi)->GetCurrSourceWordsRange();
const AlignmentInfo &alignmentInfo = (*tpi)->GetCurrTargetPhrase().GetAlignTerm();
// language model backoff stats,
if (doLMStats) {
std::vector<std::vector<unsigned int> >& lmstats = *(*tpi)->GetLMStats();
std::vector<std::vector<unsigned int> >::iterator i = lmstats.begin();
std::vector<unsigned int>::iterator acc = lmAcc.begin();
for (; i != lmstats.end(); ++i, ++acc) {
std::vector<unsigned int>::iterator j = i->begin();
lmCalls += i->size();
for (; j != i->end(); ++j) {
(*acc) += *j;
}
}
}
bool epsilon = false;
if (target == "") {
target="<EPSILON>";
epsilon = true;
droppedWords.push_back(source);
}
os << " SOURCE: " << swr << " " << source << std::endl
<< " TRANSLATED AS: " << target << std::endl
<< " WORD ALIGNED: " << alignmentInfo << std::endl;
size_t twr_i = twr.GetStartPos();
size_t swr_i = swr.GetStartPos();
if (!epsilon) {
sms << twr_i;
}
if (epsilon) {
tms << "del(" << swr_i << ")";
} else {
tms << swr_i;
}
swr_i++;
twr_i++;
for (; twr_i <= twr.GetEndPos() && twr.GetEndPos() != NOT_FOUND; twr_i++) {
sms << '-' << twr_i;
}
for (; swr_i <= swr.GetEndPos() && swr.GetEndPos() != NOT_FOUND; swr_i++) {
tms << '-' << swr_i;
}
if (!epsilon) targetMap.push_back(sms.str());
sourceMap.push_back(tms.str());
}
std::vector<std::string>::iterator si = sourceMap.begin();
std::vector<std::string>::iterator ti = targetMap.begin();
os << std::endl << "SOURCE/TARGET SPANS:";
os << std::endl << " SOURCE:";
for (; si != sourceMap.end(); ++si) {
os << " " << *si;
}
os << std::endl << " TARGET:";
for (; ti != targetMap.end(); ++ti) {
os << " " << *ti;
}
os << std::endl << std::endl;
if (doLMStats && lmCalls > 0) {
std::vector<unsigned int>::iterator acc = lmAcc.begin();
const std::vector<const StatefulFeatureFunction*> &statefulFFs = StatefulFeatureFunction::GetStatefulFeatureFunctions();
for (size_t i = 0; i < statefulFFs.size(); ++i) {
const StatefulFeatureFunction *ff = statefulFFs[i];
const LanguageModel *lm = dynamic_cast<const LanguageModel*>(ff);
if (lm) {
char buf[256];
sprintf(buf, "%.4f", (float)(*acc)/(float)lmCalls);
os << lm->GetScoreProducerDescription() <<", AVG N-GRAM LENGTH: " << buf << std::endl;
++acc;
}
}
}
if (droppedWords.size() > 0) {
std::vector<std::string>::iterator dwi = droppedWords.begin();
os << std::endl << "WORDS/PHRASES DROPPED:" << std::endl;
for (; dwi != droppedWords.end(); ++dwi) {
os << "\tdropped=" << *dwi << std::endl;
}
}
os << std::endl << "SCORES (UNWEIGHTED/WEIGHTED): ";
os << translationPath.back()->GetScoreBreakdown();
os << " weighted(TODO)";
os << std::endl;
}
void PrintTranslationAnalysis(std::ostream &os, const Moses::ChartHypothesis* hypo)
{
/*
os << endl << "TRANSLATION HYPOTHESIS DETAILS:" << endl;
queue<const Hypothesis*> translationPath;
while (hypo)
{
translationPath.push(hypo);
hypo = hypo->GetPrevHypo();
}
while (!translationPath.empty())
{
hypo = translationPath.front();
translationPath.pop();
const TranslationOption *transOpt = hypo->GetTranslationOption();
if (transOpt != NULL)
{
os << hypo->GetCurrSourceWordsRange() << " ";
for (size_t decodeStepId = 0; decodeStepId < DecodeStepTranslation::GetNumTransStep(); ++decodeStepId)
os << decodeStepId << "=" << transOpt->GetSubRangeCount(decodeStepId) << ",";
os << *transOpt << endl;
}
}
os << "END TRANSLATION" << endl;
*/
}
}
|
