File size: 5,905 Bytes
158b61b |
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 |
/***********************************************************************
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
***********************************************************************/
#ifndef moses_PhraseDictionaryMultiModelCounts_h
#define moses_PhraseDictionaryMultiModelCounts_h
#include "moses/TranslationModel/PhraseDictionaryMultiModel.h"
#include <boost/unordered_map.hpp>
#include "moses/StaticData.h"
#include "moses/TargetPhrase.h"
#include "moses/Util.h"
#include <exception>
namespace Moses
{
typedef boost::unordered_map<Word, double > lexicalMap;
typedef boost::unordered_map<Word, lexicalMap > lexicalMapJoint;
typedef std::pair<std::vector<float>, std::vector<float> > lexicalPair;
typedef std::vector<std::vector<lexicalPair> > lexicalCache;
struct multiModelCountsStats : multiModelStats {
std::vector<float> fst, ft;
};
struct multiModelCountsStatsOptimization: multiModelCountsStats {
std::vector<float> fs;
lexicalCache lexCachee2f, lexCachef2e;
size_t f;
};
struct lexicalTable {
lexicalMapJoint joint;
lexicalMap marginal;
};
double InstanceWeighting(std::vector<float> &joint_counts, std::vector<float> &marginals, std::vector<float> &multimodelweights);
double LinearInterpolationFromCounts(std::vector<float> &joint_counts, std::vector<float> &marginals, std::vector<float> &multimodelweights);
//thrown if alignment information does not match phrase pair (out-of-bound alignment points)
class AlignmentException : public std::runtime_error
{
public:
AlignmentException() : std::runtime_error("AlignmentException") { }
};
/** Implementation of a phrase table with raw counts.
*/
class PhraseDictionaryMultiModelCounts: public PhraseDictionaryMultiModel
{
#ifdef WITH_DLIB
friend class CrossEntropyCounts;
#endif
typedef std::vector< std::set<size_t> > AlignVector;
public:
PhraseDictionaryMultiModelCounts(const std::string &line);
~PhraseDictionaryMultiModelCounts();
void Load(AllOptions::ptr const& opts);
TargetPhraseCollection::shared_ptr CreateTargetPhraseCollectionCounts(const Phrase &src, std::vector<float> &fs, std::map<std::string,multiModelCountsStats*>* allStats, std::vector<std::vector<float> > &multimodelweights) const;
void CollectSufficientStats(const Phrase &src, std::vector<float> &fs, std::map<std::string,multiModelCountsStats*>* allStats) const;
float GetTargetCount(const Phrase& target, size_t modelIndex) const;
double GetLexicalProbability( Word &inner, Word &outer, const std::vector<lexicalTable*> &tables, std::vector<float> &multimodelweights ) const;
double ComputeWeightedLexicalTranslation( const Phrase &phraseS, const Phrase &phraseT, AlignVector &alignment, const std::vector<lexicalTable*> &tables, std::vector<float> &multimodelweights, bool is_input ) const;
double ComputeWeightedLexicalTranslationFromCache( std::vector<std::vector<std::pair<std::vector<float>, std::vector<float> > > > &cache, std::vector<float> &weights ) const;
std::pair<PhraseDictionaryMultiModelCounts::AlignVector,PhraseDictionaryMultiModelCounts::AlignVector> GetAlignmentsForLexWeights(const Phrase &phraseS, const Phrase &phraseT, const AlignmentInfo &alignment) const;
std::vector<std::vector<std::pair<std::vector<float>, std::vector<float> > > > CacheLexicalStats( const Phrase &phraseS, const Phrase &phraseT, AlignVector &alignment, const std::vector<lexicalTable*> &tables, bool is_input );
void FillLexicalCountsJoint(Word &wordS, Word &wordT, std::vector<float> &count, const std::vector<lexicalTable*> &tables) const;
void FillLexicalCountsMarginal(Word &wordS, std::vector<float> &count, const std::vector<lexicalTable*> &tables) const;
void LoadLexicalTable( std::string &fileName, lexicalTable* ltable);
TargetPhraseCollection::shared_ptr GetTargetPhraseCollectionLEGACY(const Phrase& src) const;
#ifdef WITH_DLIB
std::vector<float> MinimizePerplexity(std::vector<std::pair<std::string, std::string> > &phrase_pair_vector);
#endif
// functions below required by base class
virtual void InitializeForInput(ttasksptr const& ttask) {
/* Don't do anything source specific here as this object is shared between threads.*/
}
void SetParameter(const std::string& key, const std::string& value);
private:
std::vector<PhraseDictionary*> m_inverse_pd;
std::vector<lexicalTable*> m_lexTable_e2f, m_lexTable_f2e;
double (*m_combineFunction) (std::vector<float> &joint_counts, std::vector<float> &marginals, std::vector<float> &multimodelweights);
std::vector<std::string> m_lexE2FStr, m_lexF2EStr, m_targetTable;
};
#ifdef WITH_DLIB
class CrossEntropyCounts: public OptimizationObjective
{
public:
CrossEntropyCounts (
std::vector<multiModelCountsStatsOptimization*> &optimizerStats,
PhraseDictionaryMultiModelCounts * model,
size_t iFeature
) {
m_optimizerStats = optimizerStats;
m_model = model;
m_iFeature = iFeature;
}
double operator() ( const dlib::matrix<double,0,1>& arg) const;
private:
std::vector<multiModelCountsStatsOptimization*> m_optimizerStats;
PhraseDictionaryMultiModelCounts * m_model;
size_t m_iFeature;
};
#endif
} // end namespace
#endif
|