// $Id$ // vim:tabstop=2 /*********************************************************************** 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 ***********************************************************************/ #include #include #include "TranslationOptionCollection.h" #include "Sentence.h" #include "DecodeStep.h" #include "LM/Base.h" #include "FactorCollection.h" #include "InputType.h" #include "LexicalReordering.h" #include "Util.h" #include "StaticData.h" #include "DecodeStepTranslation.h" #include "DecodeGraph.h" #include "moses/FF/UnknownWordPenaltyProducer.h" #include "InputPath.h" using namespace std; namespace Moses { /** helper for pruning */ bool CompareTranslationOption(const TranslationOption *a, const TranslationOption *b) { return a->GetFutureScore() > b->GetFutureScore(); } /** constructor; since translation options are indexed by coverage span, the corresponding data structure is initialized here * This fn should be called by inherited classes */ TranslationOptionCollection::TranslationOptionCollection( InputType const& src, size_t maxNoTransOptPerCoverage, float translationOptionThreshold) : m_source(src) ,m_futureScore(src.GetSize()) ,m_maxNoTransOptPerCoverage(maxNoTransOptPerCoverage) ,m_translationOptionThreshold(translationOptionThreshold) { // create 2-d vector size_t size = src.GetSize(); for (size_t startPos = 0 ; startPos < size ; ++startPos) { m_collection.push_back( vector< TranslationOptionList >() ); size_t maxSize = size - startPos; size_t maxSizePhrase = StaticData::Instance().GetMaxPhraseLength(); maxSize = std::min(maxSize, maxSizePhrase); for (size_t endPos = 0 ; endPos < maxSize ; ++endPos) { m_collection[startPos].push_back( TranslationOptionList() ); } } } /** destructor, clears out data structures */ TranslationOptionCollection::~TranslationOptionCollection() { RemoveAllInColl(m_unksrcs); RemoveAllInColl(m_phraseDictionaryQueue); } void TranslationOptionCollection::Prune() { // quit, if max size, threshold if (m_maxNoTransOptPerCoverage == 0 && m_translationOptionThreshold == -std::numeric_limits::infinity()) return; // bookkeeping for how many options used, pruned size_t total = 0; size_t totalPruned = 0; // loop through all spans size_t size = m_source.GetSize(); for (size_t startPos = 0 ; startPos < size; ++startPos) { size_t maxSize = size - startPos; size_t maxSizePhrase = StaticData::Instance().GetMaxPhraseLength(); maxSize = std::min(maxSize, maxSizePhrase); for (size_t endPos = startPos ; endPos < startPos + maxSize ; ++endPos) { // consider list for a span TranslationOptionList &fullList = GetTranslationOptionList(startPos, endPos); total += fullList.size(); // size pruning if (m_maxNoTransOptPerCoverage > 0 && fullList.size() > m_maxNoTransOptPerCoverage) { // sort in vector nth_element(fullList.begin(), fullList.begin() + m_maxNoTransOptPerCoverage, fullList.end(), CompareTranslationOption); totalPruned += fullList.size() - m_maxNoTransOptPerCoverage; // delete the rest for (size_t i = m_maxNoTransOptPerCoverage ; i < fullList.size() ; ++i) { delete fullList.Get(i); } fullList.resize(m_maxNoTransOptPerCoverage); } // threshold pruning if (fullList.size() > 1 && m_translationOptionThreshold != -std::numeric_limits::infinity()) { // first, find the best score float bestScore = -std::numeric_limits::infinity(); for (size_t i=0; i < fullList.size() ; ++i) { if (fullList.Get(i)->GetFutureScore() > bestScore) bestScore = fullList.Get(i)->GetFutureScore(); } //std::cerr << "best score for span " << startPos << "-" << endPos << " is " << bestScore << "\n"; // then, remove items that are worse than best score + threshold for (size_t i=0; i < fullList.size() ; ++i) { if (fullList.Get(i)->GetFutureScore() < bestScore + m_translationOptionThreshold) { //std::cerr << "\tremoving item " << i << ", score " << fullList.Get(i)->GetFutureScore() << ": " << fullList.Get(i)->GetTargetPhrase() << "\n"; delete fullList.Get(i); fullList.Remove(i); total--; totalPruned++; i--; } //else //{ // std::cerr << "\tkeeping item " << i << ", score " << fullList.Get(i)->GetFutureScore() << ": " << fullList.Get(i)->GetTargetPhrase() << "\n"; //} } } // end of threshold pruning } } // end of loop through all spans VERBOSE(2," Total translation options: " << total << std::endl << "Total translation options pruned: " << totalPruned << std::endl); } /** Force a creation of a translation option where there are none for a particular source position. * ie. where a source word has not been translated, create a translation option by * 1. not observing the table limits on phrase/generation tables * 2. using the handler ProcessUnknownWord() * Call this function once translation option collection has been filled with translation options * * This function calls for unknown words is complicated by the fact it must handle different input types. * The call stack is * Base::ProcessUnknownWord() * Inherited::ProcessUnknownWord(position) * Base::ProcessOneUnknownWord() * */ void TranslationOptionCollection::ProcessUnknownWord() { const vector& decodeGraphList = StaticData::Instance().GetDecodeGraphs(); size_t size = m_source.GetSize(); // try to translation for coverage with no trans by expanding table limit for (size_t graphInd = 0 ; graphInd < decodeGraphList.size() ; graphInd++) { const DecodeGraph &decodeGraph = *decodeGraphList[graphInd]; for (size_t pos = 0 ; pos < size ; ++pos) { TranslationOptionList &fullList = GetTranslationOptionList(pos, pos); size_t numTransOpt = fullList.size(); if (numTransOpt == 0) { CreateTranslationOptionsForRange(decodeGraph, pos, pos, false, graphInd); } } } bool alwaysCreateDirectTranslationOption = StaticData::Instance().IsAlwaysCreateDirectTranslationOption(); // create unknown words for 1 word coverage where we don't have any trans options for (size_t pos = 0 ; pos < size ; ++pos) { TranslationOptionList &fullList = GetTranslationOptionList(pos, pos); if (fullList.size() == 0 || alwaysCreateDirectTranslationOption) ProcessUnknownWord(pos); } } /** special handling of ONE unknown words. Either add temporarily add word to translation table, * or drop the translation. * This function should be called by the ProcessOneUnknownWord() in the inherited class * At the moment, this unknown word handler is a bit of a hack, if copies over each factor from source * to target word, or uses the 'UNK' factor. * Ideally, this function should be in a class which can be expanded upon, for example, * to create a morphologically aware handler. * * \param sourceWord the unknown word * \param sourcePos * \param length length covered by this word (may be > 1 for lattice input) * \param inputScores a set of scores associated with unknown word (input scores from latties/CNs) */ void TranslationOptionCollection::ProcessOneUnknownWord(const Word &sourceWord,size_t sourcePos, size_t length, const Scores *inputScores) { const StaticData &staticData = StaticData::Instance(); const UnknownWordPenaltyProducer *unknownWordPenaltyProducer = staticData.GetUnknownWordPenaltyProducer(); float unknownScore = FloorScore(TransformScore(0)); // unknown word, add as trans opt FactorCollection &factorCollection = FactorCollection::Instance(); size_t isDigit = 0; const Factor *f = sourceWord[0]; // TODO hack. shouldn't know which factor is surface const StringPiece s = f->GetString(); bool isEpsilon = (s=="" || s==EPSILON); if (StaticData::Instance().GetDropUnknown()) { isDigit = s.find_first_of("0123456789"); if (isDigit == 1) isDigit = 1; else isDigit = 0; // modify the starting bitmap } Phrase* m_unksrc = new Phrase(1); m_unksrc->AddWord() = sourceWord; m_unksrcs.push_back(m_unksrc); TranslationOption *transOpt; TargetPhrase targetPhrase; targetPhrase.SetSourcePhrase(*m_unksrc); if (!(staticData.GetDropUnknown() || isEpsilon) || isDigit) { // add to dictionary Word &targetWord = targetPhrase.AddWord(); targetWord.SetIsOOV(true); for (unsigned int currFactor = 0 ; currFactor < MAX_NUM_FACTORS ; currFactor++) { FactorType factorType = static_cast(currFactor); const Factor *sourceFactor = sourceWord[currFactor]; if (sourceFactor == NULL) targetWord[factorType] = factorCollection.AddFactor(UNKNOWN_FACTOR); else targetWord[factorType] = factorCollection.AddFactor(sourceFactor->GetString()); } //create a one-to-one alignment between UNKNOWN_FACTOR and its verbatim translation targetPhrase.SetAlignmentInfo("0-0"); } else { // drop source word. create blank trans opt //targetPhrase.SetAlignment(); } targetPhrase.GetScoreBreakdown().Assign(unknownWordPenaltyProducer, unknownScore); if (inputScores != NULL) { targetPhrase.SetInputScore(*inputScores); } targetPhrase.Evaluate(*m_unksrc); transOpt = new TranslationOption(WordsRange(sourcePos, sourcePos + length - 1), targetPhrase); Add(transOpt); } /** compute future score matrix in a dynamic programming fashion. * This matrix used in search. * Call this function once translation option collection has been filled with translation options */ void TranslationOptionCollection::CalcFutureScore() { // setup the matrix (ignore lower triangle, set upper triangle to -inf size_t size = m_source.GetSize(); // the width of the matrix for(size_t row=0; row::infinity()); } } // walk all the translation options and record the cheapest option for each span for (size_t startPos = 0 ; startPos < size ; ++startPos) { size_t maxSize = m_source.GetSize() - startPos; size_t maxSizePhrase = StaticData::Instance().GetMaxPhraseLength(); maxSize = std::min(maxSize, maxSizePhrase); for (size_t endPos = startPos ; endPos < startPos + maxSize ; ++endPos) { TranslationOptionList &transOptList = GetTranslationOptionList(startPos, endPos); TranslationOptionList::const_iterator iterTransOpt; for(iterTransOpt = transOptList.begin() ; iterTransOpt != transOptList.end() ; ++iterTransOpt) { const TranslationOption &transOpt = **iterTransOpt; float score = transOpt.GetFutureScore(); if (score > m_futureScore.GetScore(startPos, endPos)) m_futureScore.SetScore(startPos, endPos, score); } } } // now fill all the cells in the strictly upper triangle // there is no way to modify the diagonal now, in the case // where no translation option covers a single-word span, // we leave the +inf in the matrix // like in chart parsing we want each cell to contain the highest score // of the full-span trOpt or the sum of scores of joining two smaller spans for(size_t colstart = 1; colstart < size ; colstart++) { for(size_t diagshift = 0; diagshift < size-colstart ; diagshift++) { size_t startPos = diagshift; size_t endPos = colstart+diagshift; for(size_t joinAt = startPos; joinAt < endPos ; joinAt++) { float joinedScore = m_futureScore.GetScore(startPos, joinAt) + m_futureScore.GetScore(joinAt+1, endPos); /* // uncomment to see the cell filling scheme TRACE_ERR( "[" < m_futureScore.GetScore(startPos, endPos)) m_futureScore.SetScore(startPos, endPos, joinedScore); } } } IFVERBOSE(3) { int total = 0; for(size_t row=0; row &decodeGraphList = StaticData::Instance().GetDecodeGraphs(); const vector &decodeGraphBackoff = StaticData::Instance().GetDecodeGraphBackoff(); // length of the sentence const size_t size = m_source.GetSize(); // loop over all decoding graphs, each generates translation options for (size_t graphInd = 0 ; graphInd < decodeGraphList.size() ; graphInd++) { if (decodeGraphList.size() > 1) { VERBOSE(3,"Creating translation options from decoding graph " << graphInd << endl); } const DecodeGraph &decodeGraph = *decodeGraphList[graphInd]; // generate phrases that start at startPos ... for (size_t startPos = 0 ; startPos < size; startPos++) { size_t maxSize = size - startPos; // don't go over end of sentence size_t maxSizePhrase = StaticData::Instance().GetMaxPhraseLength(); maxSize = std::min(maxSize, maxSizePhrase); // ... and that end at endPos for (size_t endPos = startPos ; endPos < startPos + maxSize ; endPos++) { if (graphInd > 0 && // only skip subsequent graphs decodeGraphBackoff[graphInd] != 0 && // use of backoff specified (endPos-startPos+1 >= decodeGraphBackoff[graphInd] || // size exceeds backoff limit or ... m_collection[startPos][endPos-startPos].size() > 0)) { // no phrases found so far VERBOSE(3,"No backoff to graph " << graphInd << " for span [" << startPos << ";" << endPos << "]" << endl); // do not create more options continue; } // create translation options for that range CreateTranslationOptionsForRange( decodeGraph, startPos, endPos, true, graphInd); } } } VERBOSE(2,"Translation Option Collection\n " << *this << endl); ProcessUnknownWord(); EvaluateWithSource(); // Prune Prune(); Sort(); // future score matrix CalcFutureScore(); // Cached lex reodering costs CacheLexReordering(); } void TranslationOptionCollection::CreateTranslationOptionsForRange( const DecodeGraph &decodeGraph , size_t startPos , size_t endPos , bool adhereTableLimit , size_t graphInd , InputPath &inputPath) { if ((StaticData::Instance().GetXmlInputType() != XmlExclusive) || !HasXmlOptionsOverlappingRange(startPos,endPos)) { Phrase *sourcePhrase = NULL; // can't initialise with substring, in case it's confusion network // partial trans opt stored in here PartialTranslOptColl* oldPtoc = new PartialTranslOptColl; size_t totalEarlyPruned = 0; // initial translation step list ::const_iterator iterStep = decodeGraph.begin(); const DecodeStep &decodeStep = **iterStep; const PhraseDictionary &phraseDictionary = *decodeStep.GetPhraseDictionaryFeature(); const TargetPhraseCollection *targetPhrases = inputPath.GetTargetPhrases(phraseDictionary); static_cast(decodeStep).ProcessInitialTranslation (m_source, *oldPtoc , startPos, endPos, adhereTableLimit , targetPhrases); AddInputScore(inputPath, *oldPtoc); // do rest of decode steps int indexStep = 0; for (++iterStep ; iterStep != decodeGraph.end() ; ++iterStep) { const DecodeStep *decodeStep = *iterStep; PartialTranslOptColl* newPtoc = new PartialTranslOptColl; // go thru each intermediate trans opt just created const vector& partTransOptList = oldPtoc->GetList(); vector::const_iterator iterPartialTranslOpt; for (iterPartialTranslOpt = partTransOptList.begin() ; iterPartialTranslOpt != partTransOptList.end() ; ++iterPartialTranslOpt) { TranslationOption &inputPartialTranslOpt = **iterPartialTranslOpt; if (const DecodeStepTranslation *translateStep = dynamic_cast(decodeStep) ) { const PhraseDictionary &phraseDictionary = *translateStep->GetPhraseDictionaryFeature(); const TargetPhraseCollection *targetPhrases = inputPath.GetTargetPhrases(phraseDictionary); translateStep->Process(inputPartialTranslOpt , *decodeStep , *newPtoc , this , adhereTableLimit , *sourcePhrase , targetPhrases); } else { decodeStep->Process(inputPartialTranslOpt , *decodeStep , *newPtoc , this , adhereTableLimit , *sourcePhrase); } } // last but 1 partial trans not required anymore totalEarlyPruned += newPtoc->GetPrunedCount(); delete oldPtoc; oldPtoc = newPtoc; indexStep++; } // for (++iterStep // add to fully formed translation option list PartialTranslOptColl &lastPartialTranslOptColl = *oldPtoc; const vector& partTransOptList = lastPartialTranslOptColl.GetList(); vector::const_iterator iterColl; for (iterColl = partTransOptList.begin() ; iterColl != partTransOptList.end() ; ++iterColl) { TranslationOption *transOpt = *iterColl; Add(transOpt); } lastPartialTranslOptColl.DetachAll(); totalEarlyPruned += oldPtoc->GetPrunedCount(); delete oldPtoc; // TRACE_ERR( "Early translation options pruned: " << totalEarlyPruned << endl); } // if ((StaticData::Instance().GetXmlInputType() != XmlExclusive) || !HasXmlOptionsOverlappingRange(startPos,endPos)) if (graphInd == 0 && StaticData::Instance().GetXmlInputType() != XmlPassThrough && HasXmlOptionsOverlappingRange(startPos,endPos)) { CreateXmlOptionsForRange(startPos, endPos); } } void TranslationOptionCollection::AddInputScore(const InputPath &inputPath, PartialTranslOptColl &oldPtoc) { const ScoreComponentCollection *inputScore = inputPath.GetInputScore(); if (inputScore == NULL) { return; } const std::vector &transOpts = oldPtoc.GetList(); for (size_t i = 0; i < transOpts.size(); ++i) { TranslationOption &transOpt = *transOpts[i]; ScoreComponentCollection &scores = transOpt.GetScoreBreakdown(); scores.PlusEquals(*inputScore); } } void TranslationOptionCollection::EvaluateWithSource() { const size_t size = m_source.GetSize(); for (size_t startPos = 0 ; startPos < size ; ++startPos) { size_t maxSize = m_source.GetSize() - startPos; size_t maxSizePhrase = StaticData::Instance().GetMaxPhraseLength(); maxSize = std::min(maxSize, maxSizePhrase); for (size_t endPos = startPos ; endPos < startPos + maxSize ; ++endPos) { TranslationOptionList &transOptList = GetTranslationOptionList(startPos, endPos); TranslationOptionList::const_iterator iterTransOpt; for(iterTransOpt = transOptList.begin() ; iterTransOpt != transOptList.end() ; ++iterTransOpt) { TranslationOption &transOpt = **iterTransOpt; transOpt.Evaluate(m_source); } } } } void TranslationOptionCollection::Sort() { size_t size = m_source.GetSize(); for (size_t startPos = 0 ; startPos < size; ++startPos) { size_t maxSize = size - startPos; size_t maxSizePhrase = StaticData::Instance().GetMaxPhraseLength(); maxSize = std::min(maxSize, maxSizePhrase); for (size_t endPos = startPos ; endPos < startPos + maxSize; ++endPos) { TranslationOptionList &transOptList = GetTranslationOptionList(startPos, endPos); std::sort(transOptList.begin(), transOptList.end(), CompareTranslationOption); } } } /** Check if this range overlaps with any XML options. This doesn't need to be an exact match, only an overlap. * by default, we don't support XML options. subclasses need to override this function. * called by CreateTranslationOptionsForRange() * \param startPos first position in input sentence * \param lastPos last position in input sentence * \param adhereTableLimit whether phrase & generation table limits are adhered to */ bool TranslationOptionCollection::HasXmlOptionsOverlappingRange(size_t, size_t) const { return false; //not implemented for base class } /** Populates the current Collection with XML options exactly covering the range specified. Default implementation does nothing. * called by CreateTranslationOptionsForRange() * \param startPos first position in input sentence * \param lastPos last position in input sentence */ void TranslationOptionCollection::CreateXmlOptionsForRange(size_t, size_t) { //not implemented for base class }; /** add translation option to the list * \param translationOption translation option to be added */ void TranslationOptionCollection::Add(TranslationOption *translationOption) { const WordsRange &coverage = translationOption->GetSourceWordsRange(); CHECK(coverage.GetEndPos() - coverage.GetStartPos() < m_collection[coverage.GetStartPos()].size()); m_collection[coverage.GetStartPos()][coverage.GetEndPos() - coverage.GetStartPos()].Add(translationOption); } TO_STRING_BODY(TranslationOptionCollection); std::ostream& operator<<(std::ostream& out, const TranslationOptionCollection& coll) { size_t size = coll.GetSize(); for (size_t startPos = 0 ; startPos < size ; ++startPos) { size_t maxSize = size - startPos; size_t maxSizePhrase = StaticData::Instance().GetMaxPhraseLength(); maxSize = std::min(maxSize, maxSizePhrase); for (size_t endPos = startPos ; endPos < startPos + maxSize ; ++endPos) { const TranslationOptionList& fullList = coll.GetTranslationOptionList(startPos, endPos); size_t sizeFull = fullList.size(); for (size_t i = 0; i < sizeFull; i++) { out << *fullList.Get(i) << std::endl; } } } //std::vector< std::vector< TranslationOptionList > >::const_iterator i = coll.m_collection.begin(); //size_t j = 0; //for (; i!=coll.m_collection.end(); ++i) { //out << "s[" << j++ << "].size=" << i->size() << std::endl; //} return out; } const std::vector& TranslationOptionCollection::GetUnknownSources() const { return m_unksrcs; } void TranslationOptionCollection::CacheLexReordering() { size_t size = m_source.GetSize(); const std::vector &ffs = StatefulFeatureFunction::GetStatefulFeatureFunctions(); std::vector::const_iterator iter; for (iter = ffs.begin(); iter != ffs.end(); ++iter) { const StatefulFeatureFunction &ff = **iter; if (typeid(ff) == typeid(LexicalReordering)) { const LexicalReordering &lexreordering = static_cast(ff); for (size_t startPos = 0 ; startPos < size ; startPos++) { size_t maxSize = size - startPos; size_t maxSizePhrase = StaticData::Instance().GetMaxPhraseLength(); maxSize = std::min(maxSize, maxSizePhrase); for (size_t endPos = startPos ; endPos < startPos + maxSize; endPos++) { TranslationOptionList &transOptList = GetTranslationOptionList( startPos, endPos); TranslationOptionList::iterator iterTransOpt; for(iterTransOpt = transOptList.begin() ; iterTransOpt != transOptList.end() ; ++iterTransOpt) { TranslationOption &transOpt = **iterTransOpt; //Phrase sourcePhrase = m_source.GetSubString(WordsRange(startPos,endPos)); const Phrase *sourcePhrase = transOpt.GetSourcePhrase(); if (sourcePhrase) { Scores score = lexreordering.GetProb(*sourcePhrase , transOpt.GetTargetPhrase()); if (!score.empty()) transOpt.CacheLexReorderingScores(lexreordering, score); } // if (sourcePhrase) { } // for(iterTransOpt } // for (size_t endPos = startPos ; endPos < startPos + maxSize; endPos++) { } // for (size_t startPos = 0 ; startPos < size ; startPos++) { } // if (typeid(ff) == typeid(LexicalReordering)) { } // for (iter = ffs.begin(); iter != ffs.end(); ++iter) { } //! list of trans opt for a particular span TranslationOptionList &TranslationOptionCollection::GetTranslationOptionList(size_t startPos, size_t endPos) { size_t maxSize = endPos - startPos; size_t maxSizePhrase = StaticData::Instance().GetMaxPhraseLength(); maxSize = std::min(maxSize, maxSizePhrase); CHECK(maxSize < m_collection[startPos].size()); return m_collection[startPos][maxSize]; } const TranslationOptionList &TranslationOptionCollection::GetTranslationOptionList(size_t startPos, size_t endPos) const { size_t maxSize = endPos - startPos; size_t maxSizePhrase = StaticData::Instance().GetMaxPhraseLength(); maxSize = std::min(maxSize, maxSizePhrase); CHECK(maxSize < m_collection[startPos].size()); return m_collection[startPos][maxSize]; } void TranslationOptionCollection::SetTargetPhraseFromPtMatrix() { const vector &decodeGraphList = StaticData::Instance().GetDecodeGraphs(); for (size_t graphInd = 0 ; graphInd < decodeGraphList.size() ; graphInd++) { const DecodeGraph &decodeGraph = *decodeGraphList[graphInd]; list ::const_iterator iterStep; for (iterStep = decodeGraph.begin(); iterStep != decodeGraph.end() ; ++iterStep) { const DecodeStep &decodeStep = **iterStep; const DecodeStepTranslation *transStep = dynamic_cast(&decodeStep); if (transStep) { const PhraseDictionary &phraseDictionary = *transStep->GetPhraseDictionaryFeature(); phraseDictionary.SetTargetPhraseFromPtMatrix(m_phraseDictionaryQueue); } } } } } // namespace