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
|
// $Id$
/***********************************************************************
Moses - factored phrase-based language decoder
Copyright (C) 2006 University of Edinburgh
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 Street, Fifth Floor, Boston, MA 02110-1301 USA
***********************************************************************/
#pragma once
#include <vector>
#include <algorithm>
#include "LanguageModelMultiFactor.h"
#include "LanguageModelSingleFactor.h"
#include "Phrase.h"
#include "FactorCollection.h"
/* Hacked up LM which skips any factor with string '---'
* order of chunk hardcoded to 3 (m_realNGramOrder)
*/
class LanguageModelSkip : public LanguageModelSingleFactor
{
protected:
size_t m_realNGramOrder;
LanguageModelSingleFactor *m_lmImpl;
public:
/** Constructor
* \param lmImpl SRI or IRST LM which this LM can use to load data
*/
LanguageModelSkip(LanguageModelSingleFactor *lmImpl
, bool registerScore
, ScoreIndexManager &scoreIndexManager)
: LanguageModelSingleFactor(registerScore, scoreIndexManager)
{
m_lmImpl = lmImpl;
}
~LanguageModelSkip()
{
delete m_lmImpl;
}
bool Load(const std::string &filePath
, FactorType factorType
, float weight
, size_t nGramOrder)
{
m_factorType = factorType;
m_weight = weight;
m_filePath = filePath;
m_nGramOrder = nGramOrder;
m_realNGramOrder = 3;
FactorCollection &factorCollection = FactorCollection::Instance();
m_sentenceStartArray[m_factorType] = factorCollection.AddFactor(Output, m_factorType, BOS_);
m_sentenceEndArray[m_factorType] = factorCollection.AddFactor(Output, m_factorType, EOS_);
return m_lmImpl->Load(filePath, m_factorType, weight, nGramOrder);
}
float GetValue(const std::vector<const Word*> &contextFactor, State* finalState = NULL, unsigned int* len = NULL) const
{
if (contextFactor.size() == 0)
{
return 0;
}
// only process context where last word is a word we want
const Factor *factor = (*contextFactor.back())[m_factorType];
std::string strWord = factor->GetString();
if (strWord.find("---") == 0)
return 0;
// add last word
std::vector<const Word*> chunkContext;
Word* chunkWord = new Word;
chunkWord->SetFactor(m_factorType, factor);
chunkContext.push_back(chunkWord);
// create context in reverse 'cos we skip words we don't want
for (int currPos = (int)contextFactor.size() - 2 ; currPos >= 0 && chunkContext.size() < m_realNGramOrder ; --currPos )
{
const Word &word = *contextFactor[currPos];
factor = word[m_factorType];
std::string strWord = factor->GetString();
bool skip = strWord.find("---") == 0;
if (skip)
continue;
// add word to chunked context
Word* chunkWord = new Word;
chunkWord->SetFactor(m_factorType, factor);
chunkContext.push_back(chunkWord);
}
// create context factor the right way round
std::reverse(chunkContext.begin(), chunkContext.end());
// calc score on chunked phrase
float ret = m_lmImpl->GetValue(chunkContext, finalState, len);
RemoveAllInColl(chunkContext);
return ret;
}
};
|
