bool RuleTableLoaderStandard::Load(FormatType format , const std::vector<FactorType> &input , const std::vector<FactorType> &output , const std::string &inFile , const std::vector<float> &weight , size_t /* tableLimit */ , const LMList &languageModels , const WordPenaltyProducer* wpProducer , RuleTableTrie &ruleTable) { PrintUserTime(string("Start loading text SCFG phrase table. ") + (format==MosesFormat?"Moses ":"Hiero ") + " format"); const StaticData &staticData = StaticData::Instance(); const std::string& factorDelimiter = staticData.GetFactorDelimiter(); string lineOrig; size_t count = 0; std::ostream *progress = NULL; IFVERBOSE(1) progress = &std::cerr; util::FilePiece in(inFile.c_str(), progress); // reused variables vector<float> scoreVector; StringPiece line; std::string hiero_before, hiero_after; while(true) { try { line = in.ReadLine(); } catch (const util::EndOfFileException &e) { break; } if (format == HieroFormat) { // inefficiently reformat line hiero_before.assign(line.data(), line.size()); ReformatHieroRule(hiero_before, hiero_after); line = hiero_after; } util::TokenIter<util::MultiCharacter> pipes(line, "|||"); StringPiece sourcePhraseString(*pipes); StringPiece targetPhraseString(*++pipes); StringPiece scoreString(*++pipes); StringPiece alignString(*++pipes); // TODO(bhaddow) efficiently handle default instead of parsing this string every time. StringPiece ruleCountString = ++pipes ? *pipes : StringPiece("1 1"); if (++pipes) { stringstream strme; strme << "Syntax error at " << ruleTable.GetFilePath() << ":" << count; UserMessage::Add(strme.str()); abort(); } bool isLHSEmpty = (sourcePhraseString.find_first_not_of(" \t", 0) == string::npos); if (isLHSEmpty && !staticData.IsWordDeletionEnabled()) { TRACE_ERR( ruleTable.GetFilePath() << ":" << count << ": pt entry contains empty target, skipping\n"); continue; } scoreVector.clear(); for (util::TokenIter<util::AnyCharacter, true> s(scoreString, " \t"); s; ++s) { char *err_ind; scoreVector.push_back(strtod(s->data(), &err_ind)); UTIL_THROW_IF(err_ind == s->data(), util::Exception, "Bad score " << *s << " on line " << count); } const size_t numScoreComponents = ruleTable.GetFeature()->GetNumScoreComponents(); if (scoreVector.size() != numScoreComponents) { stringstream strme; strme << "Size of scoreVector != number (" << scoreVector.size() << "!=" << numScoreComponents << ") of score components on line " << count; UserMessage::Add(strme.str()); abort(); } // parse source & find pt node // constituent labels Word sourceLHS, targetLHS; // source Phrase sourcePhrase( 0); sourcePhrase.CreateFromStringNewFormat(Input, input, sourcePhraseString, factorDelimiter, sourceLHS); // create target phrase obj TargetPhrase *targetPhrase = new TargetPhrase(Output); targetPhrase->CreateFromStringNewFormat(Output, output, targetPhraseString, factorDelimiter, targetLHS); targetPhrase->SetSourcePhrase(sourcePhrase); // rest of target phrase targetPhrase->SetAlignmentInfo(alignString, sourcePhrase); targetPhrase->SetTargetLHS(targetLHS); targetPhrase->SetRuleCount(ruleCountString, scoreVector[0]); //targetPhrase->SetDebugOutput(string("New Format pt ") + line); // component score, for n-best output std::transform(scoreVector.begin(),scoreVector.end(),scoreVector.begin(),TransformScore); std::transform(scoreVector.begin(),scoreVector.end(),scoreVector.begin(),FloorScore); targetPhrase->SetScoreChart(ruleTable.GetFeature(), scoreVector, weight, languageModels,wpProducer); TargetPhraseCollection &phraseColl = GetOrCreateTargetPhraseCollection(ruleTable, sourcePhrase, *targetPhrase, sourceLHS); phraseColl.Add(targetPhrase); count++; } // sort and prune each target phrase collection SortAndPrune(ruleTable); return true; }
void PhraseDictionaryFuzzyMatch::InitializeForInput(InputType const& inputSentence) { char dirName[] = "/tmp/moses.XXXXXX"; char *temp = mkdtemp(dirName); CHECK(temp); string dirNameStr(dirName); string inFileName(dirNameStr + "/in"); ofstream inFile(inFileName.c_str()); for (size_t i = 1; i < inputSentence.GetSize() - 1; ++i) { inFile << inputSentence.GetWord(i); } inFile << endl; inFile.close(); long translationId = inputSentence.GetTranslationId(); string ptFileName = m_FuzzyMatchWrapper->Extract(translationId, dirNameStr); // populate with rules for this sentence PhraseDictionaryNodeSCFG &rootNode = m_collection[translationId]; FormatType format = MosesFormat; // data from file InputFileStream inStream(ptFileName); // copied from class LoaderStandard PrintUserTime("Start loading fuzzy-match phrase model"); const StaticData &staticData = StaticData::Instance(); const std::string& factorDelimiter = staticData.GetFactorDelimiter(); string lineOrig; size_t count = 0; while(getline(inStream, lineOrig)) { //mgjang std add const string *line; if (format == HieroFormat) { // reformat line assert(false); //line = ReformatHieroRule(lineOrig); } else { // do nothing to format of line line = &lineOrig; } vector<string> tokens; vector<float> scoreVector; TokenizeMultiCharSeparator(tokens, *line , "|||" ); if (tokens.size() != 4 && tokens.size() != 5) { stringstream strme; strme << "Syntax error at " << ptFileName << ":" << count; UserMessage::Add(strme.str()); LOGE("[mgjang] before abort\n"); abort(); } const string &sourcePhraseString = tokens[0] , &targetPhraseString = tokens[1] , &scoreString = tokens[2] , &alignString = tokens[3]; bool isLHSEmpty = (sourcePhraseString.find_first_not_of(" \t", 0) == string::npos); if (isLHSEmpty && !staticData.IsWordDeletionEnabled()) { TRACE_ERR( ptFileName << ":" << count << ": pt entry contains empty target, skipping\n"); continue; } Tokenize<float>(scoreVector, scoreString); const size_t numScoreComponents = GetFeature()->GetNumScoreComponents(); if (scoreVector.size() != numScoreComponents) { stringstream strme; strme << "Size of scoreVector != number (" << scoreVector.size() << "!=" << numScoreComponents << ") of score components on line " << count; UserMessage::Add(strme.str()); LOGE("[mgjang] before abort\n"); abort(); } CHECK(scoreVector.size() == numScoreComponents); // parse source & find pt node // constituent labels Word sourceLHS, targetLHS; // source Phrase sourcePhrase( 0); sourcePhrase.CreateFromStringNewFormat(Input, *m_input, sourcePhraseString, factorDelimiter, sourceLHS); // create target phrase obj TargetPhrase *targetPhrase = new TargetPhrase(); targetPhrase->CreateFromStringNewFormat(Output, *m_output, targetPhraseString, factorDelimiter, targetLHS); // rest of target phrase targetPhrase->SetAlignmentInfo(alignString); targetPhrase->SetTargetLHS(targetLHS); //targetPhrase->SetDebugOutput(string("New Format pt ") + line); // component score, for n-best output std::transform(scoreVector.begin(),scoreVector.end(),scoreVector.begin(),TransformScore); std::transform(scoreVector.begin(),scoreVector.end(),scoreVector.begin(),FloorScore); targetPhrase->SetScoreChart(GetFeature(), scoreVector, *m_weight, *m_languageModels, m_wpProducer); TargetPhraseCollection &phraseColl = GetOrCreateTargetPhraseCollection(rootNode, sourcePhrase, *targetPhrase, sourceLHS); phraseColl.Add(targetPhrase); count++; if (format == HieroFormat) { // reformat line delete line; } else { // do nothing } } // sort and prune each target phrase collection SortAndPrune(rootNode); //removedirectoryrecursively(dirName); }
bool PhraseDictionaryNewFormat::Load(const std::vector<FactorType> &input , const std::vector<FactorType> &output , std::istream &inStream , const std::vector<float> &weight , size_t tableLimit , const LMList &languageModels , float weightWP) { PrintUserTime("Start loading new format pt model"); const StaticData &staticData = StaticData::Instance(); const std::string& factorDelimiter = staticData.GetFactorDelimiter(); VERBOSE(2,"PhraseDictionaryNewFormat: input=" << m_inputFactors << " output=" << m_outputFactors << std::endl); string line; size_t count = 0; while(getline(inStream, line)) { vector<string> tokens; vector<float> scoreVector; TokenizeMultiCharSeparator(tokens, line , "|||" ); if (tokens.size() != 4 && tokens.size() != 5) { stringstream strme; strme << "Syntax error at " << m_filePath << ":" << count; UserMessage::Add(strme.str()); abort(); } const string &sourcePhraseString = tokens[0] , &targetPhraseString = tokens[1] , &alignString = tokens[2] , &scoreString = tokens[3]; bool isLHSEmpty = (sourcePhraseString.find_first_not_of(" \t", 0) == string::npos); if (isLHSEmpty && !staticData.IsWordDeletionEnabled()) { TRACE_ERR( m_filePath << ":" << count << ": pt entry contains empty target, skipping\n"); continue; } Tokenize<float>(scoreVector, scoreString); if (scoreVector.size() != m_numScoreComponent) { stringstream strme; strme << "Size of scoreVector != number (" <<scoreVector.size() << "!=" <<m_numScoreComponent<<") of score components on line " << count; UserMessage::Add(strme.str()); abort(); } assert(scoreVector.size() == m_numScoreComponent); // parse source & find pt node // head word Word sourceLHS, targetLHS; // source Phrase sourcePhrase(Input); sourcePhrase.CreateFromStringNewFormat(Input, input, sourcePhraseString, factorDelimiter, sourceLHS); // create target phrase obj TargetPhrase *targetPhrase = new TargetPhrase(Output); targetPhrase->CreateFromStringNewFormat(Output, output, targetPhraseString, factorDelimiter, targetLHS); // alignment list<pair<size_t,size_t> > alignmentInfo; CreateAlignmentInfo(alignmentInfo, alignString); // rest of target phrase targetPhrase->SetAlignmentInfo(alignmentInfo); targetPhrase->SetTargetLHS(targetLHS); //targetPhrase->SetDebugOutput(string("New Format pt ") + line); // component score, for n-best output std::transform(scoreVector.begin(),scoreVector.end(),scoreVector.begin(),TransformScore); std::transform(scoreVector.begin(),scoreVector.end(),scoreVector.begin(),FloorScore); targetPhrase->SetScoreChart(GetFeature(), scoreVector, weight, languageModels); // count info for backoff if (tokens.size() >= 6) targetPhrase->CreateCountInfo(tokens[5]); TargetPhraseCollection &phraseColl = GetOrCreateTargetPhraseCollection(sourcePhrase, *targetPhrase); AddEquivPhrase(phraseColl, targetPhrase); count++; } // cleanup cache // sort each target phrase collection m_collection.Sort(m_tableLimit); return true; }
bool RuleTableLoaderStandard::Load(FormatType format , const std::vector<FactorType> &input , const std::vector<FactorType> &output , std::istream &inStream , const std::vector<float> &weight , size_t /* tableLimit */ , const LMList &languageModels , const WordPenaltyProducer* wpProducer , RuleTableTrie &ruleTable) { PrintUserTime(string("Start loading text SCFG phrase table. ") + (format==MosesFormat?"Moses ":"Hiero ") + " format"); const StaticData &staticData = StaticData::Instance(); const std::string& factorDelimiter = staticData.GetFactorDelimiter(); string lineOrig; size_t count = 0; while(getline(inStream, lineOrig)) { const string *line; if (format == HieroFormat) { // reformat line line = ReformatHieroRule(lineOrig); } else { // do nothing to format of line line = &lineOrig; } vector<string> tokens; vector<float> scoreVector; TokenizeMultiCharSeparator(tokens, *line , "|||" ); if (tokens.size() != 4 && tokens.size() != 5) { stringstream strme; strme << "Syntax error at " << ruleTable.GetFilePath() << ":" << count; UserMessage::Add(strme.str()); abort(); } const string &sourcePhraseString = tokens[0] , &targetPhraseString = tokens[1] , &scoreString = tokens[2] , &alignString = tokens[3]; bool isLHSEmpty = (sourcePhraseString.find_first_not_of(" \t", 0) == string::npos); if (isLHSEmpty && !staticData.IsWordDeletionEnabled()) { TRACE_ERR( ruleTable.GetFilePath() << ":" << count << ": pt entry contains empty target, skipping\n"); continue; } Tokenize<float>(scoreVector, scoreString); const size_t numScoreComponents = ruleTable.GetFeature()->GetNumScoreComponents(); if (scoreVector.size() != numScoreComponents) { stringstream strme; strme << "Size of scoreVector != number (" << scoreVector.size() << "!=" << numScoreComponents << ") of score components on line " << count; UserMessage::Add(strme.str()); abort(); } CHECK(scoreVector.size() == numScoreComponents); // parse source & find pt node // constituent labels Word sourceLHS, targetLHS; // source Phrase sourcePhrase( 0); sourcePhrase.CreateFromStringNewFormat(Input, input, sourcePhraseString, factorDelimiter, sourceLHS); // create target phrase obj TargetPhrase *targetPhrase = new TargetPhrase(Output); targetPhrase->CreateFromStringNewFormat(Output, output, targetPhraseString, factorDelimiter, targetLHS); // rest of target phrase targetPhrase->SetAlignmentInfo(alignString); targetPhrase->SetTargetLHS(targetLHS); //targetPhrase->SetDebugOutput(string("New Format pt ") + line); // component score, for n-best output std::transform(scoreVector.begin(),scoreVector.end(),scoreVector.begin(),TransformScore); std::transform(scoreVector.begin(),scoreVector.end(),scoreVector.begin(),FloorScore); targetPhrase->SetScoreChart(ruleTable.GetFeature(), scoreVector, weight, languageModels,wpProducer); TargetPhraseCollection &phraseColl = GetOrCreateTargetPhraseCollection(ruleTable, sourcePhrase, *targetPhrase, sourceLHS); phraseColl.Add(targetPhrase); count++; if (format == HieroFormat) { // reformat line delete line; } else { // do nothing } } // sort and prune each target phrase collection SortAndPrune(ruleTable); return true; }