C++ moses::OutputFileStream类代码示例

void writeCountOfCounts( const string &fileNameCountOfCounts )
{
  // open file
	Moses::OutputFileStream countOfCountsFile;
	bool success = countOfCountsFile.Open(fileNameCountOfCounts.c_str());
	if (!success) {
		cerr << "ERROR: could not open count-of-counts file "
				 << fileNameCountOfCounts << endl;
    return;
	}

  // Kneser-Ney needs the total number of phrase pairs
  countOfCountsFile << totalDistinct << endl;

  // write out counts
  for(int i=1; i<=COC_MAX; i++) {
    countOfCountsFile << countOfCounts[ i ] << endl;
  }
	countOfCountsFile.Close();
}

void processFiles( const std::string& fileNameDirect, 
                   const std::string& fileNameIndirect, 
                   const std::string& fileNameConsolidated, 
                   const std::string& fileNameCountOfCounts, 
                   const std::string& fileNameSourceLabelSet, 
                   const std::string& fileNamePartsOfSpeechVocabulary )
{
  if (goodTuringFlag || kneserNeyFlag)
    loadCountOfCounts( fileNameCountOfCounts );

  // open input files
  Moses::InputFileStream fileDirect(fileNameDirect);
  UTIL_THROW_IF2(fileDirect.fail(), "could not open phrase table file " << fileNameDirect);
  Moses::InputFileStream fileIndirect(fileNameIndirect);
  UTIL_THROW_IF2(fileIndirect.fail(), "could not open phrase table file " << fileNameIndirect);

  // open output file: consolidated phrase table
  Moses::OutputFileStream fileConsolidated;
  bool success = fileConsolidated.Open(fileNameConsolidated);
  UTIL_THROW_IF2(!success, "could not open output file " << fileNameConsolidated);

  // create properties consolidator
  // (in case any additional phrase property requires further processing)
  MosesTraining::PropertiesConsolidator propertiesConsolidator = MosesTraining::PropertiesConsolidator();
  if (sourceLabelsFlag) {
    propertiesConsolidator.ActivateSourceLabelsProcessing(fileNameSourceLabelSet);
  }
  if (partsOfSpeechFlag) {
    propertiesConsolidator.ActivatePartsOfSpeechProcessing(fileNamePartsOfSpeechVocabulary);
  }

  // loop through all extracted phrase translations
  int i=0;
  while(true) {
    i++;
    if (i%100000 == 0) std::cerr << "." << std::flush;

    std::vector< std::string > itemDirect, itemIndirect;
    if (! getLine(fileIndirect, itemIndirect) ||
        ! getLine(fileDirect, itemDirect))
      break;

    // direct: target source alignment probabilities
    // indirect: source target probabilities

    // consistency checks
    UTIL_THROW_IF2(itemDirect[0].compare( itemIndirect[0] ) != 0, 
                   "target phrase does not match in line " << i << ": '" << itemDirect[0] << "' != '" << itemIndirect[0] << "'");
    UTIL_THROW_IF2(itemDirect[1].compare( itemIndirect[1] ) != 0, 
                   "source phrase does not match in line " << i << ": '" << itemDirect[1] << "' != '" << itemIndirect[1] << "'");

    // SCORES ...
    std::string directScores, directSparseScores, indirectScores, indirectSparseScores;
    breakdownCoreAndSparse( itemDirect[3], directScores, directSparseScores );
    breakdownCoreAndSparse( itemIndirect[3], indirectScores, indirectSparseScores );

    std::vector<std::string> directCounts;
    Moses::Tokenize( directCounts, itemDirect[4] );
    std::vector<std::string> indirectCounts;
    Moses::Tokenize( indirectCounts, itemIndirect[4] );
    float countF = Moses::Scan<float>(directCounts[0]);
    float countE = Moses::Scan<float>(indirectCounts[0]);
    float countEF = Moses::Scan<float>(indirectCounts[1]);
    float n1_F, n1_E;
    if (kneserNeyFlag) {
      n1_F = Moses::Scan<float>(directCounts[2]);
      n1_E = Moses::Scan<float>(indirectCounts[2]);
    }

    // Good Turing discounting
    float adjustedCountEF = countEF;
    if (goodTuringFlag && countEF+0.99999 < goodTuringDiscount.size()-1)
      adjustedCountEF *= goodTuringDiscount[(int)(countEF+0.99998)];
    float adjustedCountEF_indirect = adjustedCountEF;

    // Kneser Ney discounting [Foster et al, 2006]
    if (kneserNeyFlag) {
      float D = kneserNey_D3;
      if (countEF < 2) D = kneserNey_D1;
      else if (countEF < 3) D = kneserNey_D2;
      if (D > countEF) D = countEF - 0.01; // sanity constraint

      float p_b_E = n1_E / totalCount; // target phrase prob based on distinct
      float alpha_F = D * n1_F / countF; // available mass
      adjustedCountEF = countEF - D + countF * alpha_F * p_b_E;

      // for indirect
      float p_b_F = n1_F / totalCount; // target phrase prob based on distinct
      float alpha_E = D * n1_E / countE; // available mass
      adjustedCountEF_indirect = countEF - D + countE * alpha_E * p_b_F;
    }

    // drop due to MinScore thresholding
    if ((minScore0 > 0 && adjustedCountEF_indirect/countE < minScore0) ||
        (minScore2 > 0 && adjustedCountEF         /countF < minScore2)) {
      continue;
    }

    // output phrase pair
    fileConsolidated << itemDirect[0] << " ||| ";
//.........这里部分代码省略.........

void processFiles( char* fileNameDirect, char* fileNameIndirect, char* fileNameConsolidated, char* fileNameCountOfCounts )
{
  if (goodTuringFlag || kneserNeyFlag)
    loadCountOfCounts( fileNameCountOfCounts );

  // open input files
  Moses::InputFileStream fileDirect(fileNameDirect);
  Moses::InputFileStream fileIndirect(fileNameIndirect);

  if (fileDirect.fail()) {
    cerr << "ERROR: could not open phrase table file " << fileNameDirect << endl;
    exit(1);
  }
  istream &fileDirectP = fileDirect;

  if (fileIndirect.fail()) {
    cerr << "ERROR: could not open phrase table file " << fileNameIndirect << endl;
    exit(1);
  }
  istream &fileIndirectP = fileIndirect;

  // open output file: consolidated phrase table
  Moses::OutputFileStream fileConsolidated;
  bool success = fileConsolidated.Open(fileNameConsolidated);
  if (!success) {
    cerr << "ERROR: could not open output file " << fileNameConsolidated << endl;
    exit(1);
  }

  // loop through all extracted phrase translations
  int i=0;
  while(true) {
    i++;
    if (i%100000 == 0) cerr << "." << flush;

    vector< string > itemDirect, itemIndirect;
    if (! getLine(fileIndirectP,itemIndirect) ||
        ! getLine(fileDirectP,  itemDirect  ))
      break;

    // direct: target source alignment probabilities
    // indirect: source target probabilities

    // consistency checks
    if (itemDirect[0].compare( itemIndirect[0] ) != 0) {
      cerr << "ERROR: target phrase does not match in line " << i << ": '"
           << itemDirect[0] << "' != '" << itemIndirect[0] << "'" << endl;
      exit(1);
    }

    if (itemDirect[1].compare( itemIndirect[1] ) != 0) {
      cerr << "ERROR: source phrase does not match in line " << i << ": '"
           << itemDirect[1] << "' != '" << itemIndirect[1] << "'" << endl;
      exit(1);
    }

    // output hierarchical phrase pair (with separated labels)
    fileConsolidated << itemDirect[0] << " ||| " << itemDirect[1] << " |||";

    // SCORES ...
    string directScores, directSparseScores, indirectScores, indirectSparseScores;
    breakdownCoreAndSparse( itemDirect[3], directScores, directSparseScores );
    breakdownCoreAndSparse( itemIndirect[3], indirectScores, indirectSparseScores );

    vector<string> directCounts = tokenize(itemDirect[4].c_str());
    vector<string> indirectCounts = tokenize(itemIndirect[4].c_str());
    float countF = atof(directCounts[0].c_str());
    float countE = atof(indirectCounts[0].c_str());
    float countEF = atof(indirectCounts[1].c_str());
    float n1_F, n1_E;
    if (kneserNeyFlag) {
      n1_F = atof(directCounts[2].c_str());
      n1_E = atof(indirectCounts[2].c_str());
    }

    // Good Turing discounting
    float adjustedCountEF = countEF;
    if (goodTuringFlag && countEF+0.99999 < goodTuringDiscount.size()-1)
      adjustedCountEF *= goodTuringDiscount[(int)(countEF+0.99998)];
    float adjustedCountEF_indirect = adjustedCountEF;

    // Kneser Ney discounting [Foster et al, 2006]
    if (kneserNeyFlag) {
      float D = kneserNey_D3;
      if (countEF < 2) D = kneserNey_D1;
      else if (countEF < 3) D = kneserNey_D2;
      if (D > countEF) D = countEF - 0.01; // sanity constraint

      float p_b_E = n1_E / totalCount; // target phrase prob based on distinct
      float alpha_F = D * n1_F / countF; // available mass
      adjustedCountEF = countEF - D + countF * alpha_F * p_b_E;

      // for indirect
      float p_b_F = n1_F / totalCount; // target phrase prob based on distinct
      float alpha_E = D * n1_E / countE; // available mass
      adjustedCountEF_indirect = countEF - D + countE * alpha_E * p_b_F;
    }

    // prob indirect
    if (!onlyDirectFlag) {
//.........这里部分代码省略.........

//.........这里部分代码省略.........
#ifdef WITH_THREADS
      thread_count = atoi(argv[++i]);
#else
      cerr << "thread support not compiled in." << '\n';
      exit(1);
#endif
    } else if (strcmp(argv[i], "--SentenceOffset") == 0) {
      if (i+1 >= argc || argv[i+1][0] < '0' || argv[i+1][0] > '9') {
        cerr << "extract: syntax error, used switch --SentenceOffset without a number" << endl;
        exit(1);
      }
      sentenceOffset = atoi(argv[++i]);
    } else if (strcmp(argv[i],"--BoundaryRules") == 0) {
      options.boundaryRules = true;
    } else {
      cerr << "extract: syntax error, unknown option '" << string(argv[i]) << "'\n";
      exit(1);
    }
  }

  cerr << "extracting hierarchical rules" << endl;

  // open input files
  Moses::InputFileStream tFile(fileNameT);
  Moses::InputFileStream sFile(fileNameS);
  Moses::InputFileStream aFile(fileNameA);

  istream *tFileP = &tFile;
  istream *sFileP = &sFile;
  istream *aFileP = &aFile;

  // open output files
  string fileNameExtractInv = fileNameExtract + ".inv" + (options.gzOutput?".gz":"");
  Moses::OutputFileStream extractFile;
  Moses::OutputFileStream extractFileInv;
  Moses::OutputFileStream extractFileContext;
  Moses::OutputFileStream extractFileContextInv;
  extractFile.Open((fileNameExtract  + (options.gzOutput?".gz":"")).c_str());
  if (!options.onlyDirectFlag)
    extractFileInv.Open(fileNameExtractInv.c_str());

  if (options.flexScoreFlag) {
    string fileNameExtractContext = fileNameExtract + ".context" + (options.gzOutput?".gz":"");
    extractFileContext.Open(fileNameExtractContext.c_str());
    if (!options.onlyDirectFlag) {
      string fileNameExtractContextInv = fileNameExtract + ".context.inv" + (options.gzOutput?".gz":"");
      extractFileContextInv.Open(fileNameExtractContextInv.c_str());
    }
  }

  // stats on labels for glue grammar and unknown word label probabilities
  set< string > targetLabelCollection, sourceLabelCollection;
  map< string, int > targetTopLabelCollection, sourceTopLabelCollection;

  // loop through all sentence pairs
  size_t i=sentenceOffset;
  string targetString, sourceString, alignmentString;

  while(getline(*tFileP, targetString)) {
    i++;

    getline(*sFileP, sourceString);
    getline(*aFileP, alignmentString);

    if (i%1000 == 0) cerr << i << " " << flush;

//.........这里部分代码省略.........
    //else if (strcmp(argv[i],"--ZipFiles") == 0) {
    //  zipFiles = true;
    //}
    // if an source phrase is paired with two target phrases, then count(t|s) = 0.5
    else if (strcmp(argv[i],"--NoFractionalCounting") == 0) {
      options.fractionalCounting = false;
    } else if (strcmp(argv[i],"--OutputNTLengths") == 0) {
      options.outputNTLengths = true;
#ifdef WITH_THREADS
    } else if (strcmp(argv[i],"-threads") == 0 || 
               strcmp(argv[i],"--threads") == 0 ||
               strcmp(argv[i],"--Threads") == 0) {
      thread_count = atoi(argv[++i]);
#endif
    } else {
      cerr << "extract: syntax error, unknown option '" << string(argv[i]) << "'\n";
      exit(1);
    }
  }

  cerr << "extracting hierarchical rules" << endl;

  // open input files
  Moses::InputFileStream tFile(fileNameT);
  Moses::InputFileStream sFile(fileNameS);
  Moses::InputFileStream aFile(fileNameA);

  istream *tFileP = &tFile;
  istream *sFileP = &sFile;
  istream *aFileP = &aFile;

  // open output files
  string fileNameExtractInv = fileNameExtract + ".inv" + (options.gzOutput?".gz":"");
  Moses::OutputFileStream extractFile;
  Moses::OutputFileStream extractFileInv;
  extractFile.Open((fileNameExtract  + (options.gzOutput?".gz":"")).c_str());
  if (!options.onlyDirectFlag)
    extractFileInv.Open(fileNameExtractInv.c_str());

  // output into file
  Moses::OutputCollector* extractCollector = new Moses::OutputCollector(&extractFile);
  Moses::OutputCollector* extractCollectorInv = new Moses::OutputCollector(&extractFileInv);

  // stats on labels for glue grammar and unknown word label probabilities
  set< string > targetLabelCollection, sourceLabelCollection;
  map< string, int > targetTopLabelCollection, sourceTopLabelCollection;

#ifdef WITH_THREADS
  // set up thread pool
  Moses::ThreadPool pool(thread_count);
  pool.SetQueueLimit(1000);
#endif

  // loop through all sentence pairs
  size_t i=0;
  while(true) {
    i++;
    if (i%1000 == 0) cerr << "." << flush;
    if (i%10000 == 0) cerr << ":" << flush;
    if (i%100000 == 0) cerr << "!" << flush;
    char targetString[LINE_MAX_LENGTH];
    char sourceString[LINE_MAX_LENGTH];
    char alignmentString[LINE_MAX_LENGTH];
    SAFE_GETLINE((*tFileP), targetString, LINE_MAX_LENGTH, '\n', __FILE__);
    if (tFileP->eof()) break;
    SAFE_GETLINE((*sFileP), sourceString, LINE_MAX_LENGTH, '\n', __FILE__);

//.........这里部分代码省略.........
    }
		else if (strcmp(argv[i], "--GZOutput") == 0) {
      global->gzOutput = true;
    }
		else if (strcmp(argv[i],"--MaxSpan") == 0) {
		  // ignore
      ++i;
		}
    else if (strcmp(argv[i],"--SentenceOffset") == 0) {
      if (i+1 >= argc || argv[i+1][0] < '0' || argv[i+1][0] > '9') {
        cerr << "extract: syntax error, used switch --SentenceOffset without a number" << endl;
        exit(1);
      }
      sentenceOffset = atoi(argv[++i]);
    }
    else {
      cerr << "extract: syntax error, unknown option '" << string(argv[i]) << "'\n";
      exit(1);
    }
  }


	// open input files
	Moses::InputFileStream tFile(fileNameT);
	Moses::InputFileStream sFile(fileNameS);
	Moses::InputFileStream aFile(fileNameA);

	// open output files
  string fileNameExtractInv = fileNameExtract + ".inv";
  if (global->gzOutput) {
    fileNameExtract += ".gz";
    fileNameExtractInv += ".gz";
  }

  Moses::OutputFileStream extractFile;
  Moses::OutputFileStream extractFileInv;
  extractFile.Open(fileNameExtract.c_str());
  extractFileInv.Open(fileNameExtractInv.c_str());
  
  
	// loop through all sentence pairs
  int i = sentenceOffset;
  while(true) {
    i++;

    if (i % 1000 == 0) {
      cerr << i << " " << flush;
    }

    string targetString;
    string sourceString;
    string alignmentString;
		
		bool ok = getline(tFile, targetString);
		if (!ok)
			break;
		getline(sFile, sourceString);
		getline(aFile, alignmentString);
    
		//cerr << endl << targetString << endl << sourceString << endl << alignmentString << endl;

		//time_t currTime = time(NULL);
		//cerr << "A " << (currTime - starttime) << endl;

    SentenceAlignment sentencePair;
    if (sentencePair.Create( targetString, sourceString, alignmentString, i, *global )) 
		{			
			//cerr << sentence.sourceTree << endl;
			//cerr << sentence.targetTree << endl;

			sentencePair.FindTunnels(*g_global);
			//cerr << "C " << (time(NULL) - starttime) << endl;
			//cerr << sentencePair << endl;
			
			sentencePair.CreateLattice(*g_global);
			//cerr << "D " << (time(NULL) - starttime) << endl;
			//cerr << sentencePair << endl;

			sentencePair.CreateRules(*g_global);
			//cerr << "E " << (time(NULL) - starttime) << endl;

			//cerr << sentence.lattice->GetRules().GetSize() << endl;
			sentencePair.GetLattice().GetRules().Output(extractFile);
      sentencePair.GetLattice().GetRules().OutputInv(extractFileInv);
    }
  }
	
  tFile.Close();
  sFile.Close();
  aFile.Close();

  extractFile.Close();
  extractFileInv.Close();

  if (global->glueGrammarFlag) {
    writeGlueGrammar(fileNameGlueGrammar, *global, targetLabelCollection, targetTopLabelCollection);
  }

  delete global;
}

//.........这里部分代码省略.........
  // load domain information
  if (domainFlag) {
    if (inverseFlag) {
      domainFlag = false;
      includeSentenceIdFlag = false;
    }
    else {
      domain = new Domain;
      domain->load( fileNameDomain );
    }
  }

  // compute count of counts for Good Turing discounting
  if (goodTuringFlag || kneserNeyFlag) {
    for(int i=1; i<=COC_MAX; i++) countOfCounts[i] = 0;
  }

  // sorted phrase extraction file
  Moses::InputFileStream extractFile(fileNameExtract);

  if (extractFile.fail()) {
    cerr << "ERROR: could not open extract file " << fileNameExtract << endl;
    exit(1);
  }
  istream &extractFileP = extractFile;

  // output file: phrase translation table
	ostream *phraseTableFile;

	if (fileNamePhraseTable == "-") {
		phraseTableFile = &cout;
	}
	else {
		Moses::OutputFileStream *outputFile = new Moses::OutputFileStream();
		bool success = outputFile->Open(fileNamePhraseTable);
		if (!success) {
			cerr << "ERROR: could not open file phrase table file "
					 << fileNamePhraseTable << endl;
			exit(1);
		}
		phraseTableFile = outputFile;
	}
	
  // loop through all extracted phrase translations
  float lastCount = 0.0f;
  float lastPcfgSum = 0.0f;
  vector< PhraseAlignment > phrasePairsWithSameF;
  bool isSingleton = true;
  int i=0;
  char line[LINE_MAX_LENGTH],lastLine[LINE_MAX_LENGTH];
  lastLine[0] = '\0';
  PhraseAlignment *lastPhrasePair = NULL;
  while(true) {
    if (extractFileP.eof()) break;
    if (++i % 100000 == 0) cerr << "." << flush;
    SAFE_GETLINE((extractFileP), line, LINE_MAX_LENGTH, '\n', __FILE__);
    if (extractFileP.eof())	break;

    // identical to last line? just add count
    if (strcmp(line,lastLine) == 0) {
      lastPhrasePair->count += lastCount;
      lastPhrasePair->pcfgSum += lastPcfgSum;
      continue;
    }
    strcpy( lastLine, line );

int main(int argc, char* argv[])
{
  cerr	<< "PhraseExtract v1.4, written by Philipp Koehn\n"
        << "phrase extraction from an aligned parallel corpus\n";

  if (argc < 6) {
    cerr << "syntax: extract en de align extract max-length [orientation [ --model [wbe|phrase|hier]-[msd|mslr|mono] ] ";
    cerr<<"| --OnlyOutputSpanInfo | --NoTTable | --GZOutput | --IncludeSentenceId | --SentenceOffset n | --InstanceWeights filename ]\n";
    exit(1);
  }

  Moses::OutputFileStream extractFileOrientation;
  const char* const &fileNameE = argv[1];
  const char* const &fileNameF = argv[2];
  const char* const &fileNameA = argv[3];
  const string fileNameExtract = string(argv[4]);
  PhraseExtractionOptions options(atoi(argv[5]));

  for(int i=6; i<argc; i++) {
    if (strcmp(argv[i],"--OnlyOutputSpanInfo") == 0) {
      options.initOnlyOutputSpanInfo(true);
    } else if (strcmp(argv[i],"orientation") == 0 || strcmp(argv[i],"--Orientation") == 0) {
      options.initOrientationFlag(true);
    } else if (strcmp(argv[i],"--FlexibilityScore") == 0) {
      options.initFlexScoreFlag(true);
    } else if (strcmp(argv[i],"--NoTTable") == 0) {
      options.initTranslationFlag(false);
    } else if (strcmp(argv[i], "--IncludeSentenceId") == 0) {
      options.initIncludeSentenceIdFlag(true);
    } else if (strcmp(argv[i], "--SentenceOffset") == 0) {
      if (i+1 >= argc || argv[i+1][0] < '0' || argv[i+1][0] > '9') {
        cerr << "extract: syntax error, used switch --SentenceOffset without a number" << endl;
        exit(1);
      }
      sentenceOffset = atoi(argv[++i]);
    } else if (strcmp(argv[i], "--GZOutput") == 0) {
      options.initGzOutput(true);
    } else if (strcmp(argv[i], "--InstanceWeights") == 0) {
      if (i+1 >= argc) {
        cerr << "extract: syntax error, used switch --InstanceWeights without file name" << endl;
        exit(1);
      }
      options.initInstanceWeightsFile(argv[++i]);
    } else if (strcmp(argv[i], "--Debug") == 0) {
    	options.debug = true;
    } else if (strcmp(argv[i], "--MinPhraseLength") == 0) {
    	options.minPhraseLength = atoi(argv[++i]);
    } else if (strcmp(argv[i], "--Separator") == 0) {
    	options.separator = argv[++i];
    } else if(strcmp(argv[i],"--model") == 0) {
      if (i+1 >= argc) {
        cerr << "extract: syntax error, no model's information provided to the option --model " << endl;
        exit(1);
      }
      char*  modelParams = argv[++i];
      char*  modelName = strtok(modelParams, "-");
      char*  modelType = strtok(NULL, "-");

      // REO_MODEL_TYPE intModelType;

      if(strcmp(modelName, "wbe") == 0) {
        options.initWordModel(true);
        if(strcmp(modelType, "msd") == 0)
          options.initWordType(REO_MSD);
        else if(strcmp(modelType, "mslr") == 0)
          options.initWordType(REO_MSLR);
        else if(strcmp(modelType, "mono") == 0 || strcmp(modelType, "monotonicity") == 0)
          options.initWordType(REO_MONO);
        else {
          cerr << "extract: syntax error, unknown reordering model type: " << modelType << endl;
          exit(1);
        }
      } else if(strcmp(modelName, "phrase") == 0) {
        options.initPhraseModel(true);
        if(strcmp(modelType, "msd") == 0)
          options.initPhraseType(REO_MSD);
        else if(strcmp(modelType, "mslr") == 0)
          options.initPhraseType(REO_MSLR);
        else if(strcmp(modelType, "mono") == 0 || strcmp(modelType, "monotonicity") == 0)
          options.initPhraseType(REO_MONO);
        else {
          cerr << "extract: syntax error, unknown reordering model type: " << modelType << endl;
          exit(1);
        }
      } else if(strcmp(modelName, "hier") == 0) {
        options.initHierModel(true);
        if(strcmp(modelType, "msd") == 0)
          options.initHierType(REO_MSD);
        else if(strcmp(modelType, "mslr") == 0)
          options.initHierType(REO_MSLR);
        else if(strcmp(modelType, "mono") == 0 || strcmp(modelType, "monotonicity") == 0)
          options.initHierType(REO_MONO);
        else {
          cerr << "extract: syntax error, unknown reordering model type: " << modelType << endl;
          exit(1);
        }
      } else {
        cerr << "extract: syntax error, unknown reordering model: " << modelName << endl;
        exit(1);
      }
//.........这里部分代码省略.........

int main(int argc, char* argv[])
{
  cerr << "Starting..." << endl;

  char* &fileNameDirect = argv[1];
  Moses::InputFileStream fileDirect(fileNameDirect);


  //fileDirect.open(fileNameDirect);
  if (fileDirect.fail()) {
    cerr << "ERROR: could not open extract file " << fileNameDirect << endl;
    exit(1);
  }
  istream &fileDirectP = fileDirect;

  char* &fileNameConsolidated = argv[2];
  ostream *fileConsolidated;

  if (strcmp(fileNameConsolidated, "-") == 0) {
    fileConsolidated = &cout;
  } else {
    Moses::OutputFileStream *outputFile = new Moses::OutputFileStream();
    bool success = outputFile->Open(fileNameConsolidated);
    if (!success) {
      cerr << "ERROR: could not open file phrase table file "
           << fileNameConsolidated << endl;
      exit(1);
    }
    fileConsolidated = outputFile;
  }

  int i=0;
  while(true) {
    i++;
    if (i%1000 == 0) cerr << "." << flush;
    if (i%10000 == 0) cerr << ":" << flush;
    if (i%100000 == 0) cerr << "!" << flush;

    vector< string > itemDirect;
    if (! getLine(fileDirectP,  itemDirect  ))
      break;

    (*fileConsolidated) << itemDirect[0] << " ||| " << itemDirect[1] << " ||| ";

    // output alignment and probabilities
    (*fileConsolidated)	<< itemDirect[2]						// prob direct
                        << " 2.718" // phrase count feature
                        << " ||| " << itemDirect[3];	// alignment

    // counts
    (*fileConsolidated) << "||| 0 " << itemDirect[4]; // indirect
    (*fileConsolidated) << endl;

  }

  fileConsolidated->flush();
  if (fileConsolidated != &cout) {
    delete fileConsolidated;
  }

  cerr << "Finished" << endl;
}

int main(int argc, char* argv[])
{
  cerr	<< "PhraseExtract v1.4, written by Philipp Koehn\n"
        << "phrase extraction from an aligned parallel corpus\n";

  if (argc < 6) {
    cerr << "syntax: extract en de align extract max-length [orientation [ --model [wbe|phrase|hier]-[msd|mslr|mono] ] | --OnlyOutputSpanInfo | --NoTTable | --SentenceId]\n";
    exit(1);
  }
  char* &fileNameE = argv[1];
  char* &fileNameF = argv[2];
  char* &fileNameA = argv[3];
  string fileNameExtract = string(argv[4]);
  maxPhraseLength = atoi(argv[5]);

  for(int i=6; i<argc; i++) {
    if (strcmp(argv[i],"--OnlyOutputSpanInfo") == 0) {
      onlyOutputSpanInfo = true;
    } else if (strcmp(argv[i],"orientation") == 0 || strcmp(argv[i],"--Orientation") == 0) {
      orientationFlag = true;
    } else if (strcmp(argv[i],"--NoTTable") == 0) {
      translationFlag = false;
    } else if (strcmp(argv[i], "--SentenceId") == 0) {
      sentenceIdFlag = true;  
    } else if (strcmp(argv[i], "--GZOutput") == 0) {
      gzOutput = true;  
    } else if(strcmp(argv[i],"--model") == 0) {
      if (i+1 >= argc) {
        cerr << "extract: syntax error, no model's information provided to the option --model " << endl;
        exit(1);
      }
      char* modelParams = argv[++i];
      char* modelName = strtok(modelParams, "-");
      char* modelType = strtok(NULL, "-");

      REO_MODEL_TYPE intModelType;

      if(strcmp(modelName, "wbe") == 0) {
        wordModel = true;
        if(strcmp(modelType, "msd") == 0)
          wordType = REO_MSD;
        else if(strcmp(modelType, "mslr") == 0)
          wordType = REO_MSLR;
        else if(strcmp(modelType, "mono") == 0 || strcmp(modelType, "monotonicity") == 0)
          wordType = REO_MONO;
        else {
          cerr << "extract: syntax error, unknown reordering model type: " << modelType << endl;
          exit(1);
        }
      } else if(strcmp(modelName, "phrase") == 0) {
        phraseModel = true;
        if(strcmp(modelType, "msd") == 0)
          phraseType = REO_MSD;
        else if(strcmp(modelType, "mslr") == 0)
          phraseType = REO_MSLR;
        else if(strcmp(modelType, "mono") == 0 || strcmp(modelType, "monotonicity") == 0)
          phraseType = REO_MONO;
        else {
          cerr << "extract: syntax error, unknown reordering model type: " << modelType << endl;
          exit(1);
        }
      } else if(strcmp(modelName, "hier") == 0) {
        hierModel = true;
        if(strcmp(modelType, "msd") == 0)
          hierType = REO_MSD;
        else if(strcmp(modelType, "mslr") == 0)
          hierType = REO_MSLR;
        else if(strcmp(modelType, "mono") == 0 || strcmp(modelType, "monotonicity") == 0)
          hierType = REO_MONO;
        else {
          cerr << "extract: syntax error, unknown reordering model type: " << modelType << endl;
          exit(1);
        }
      } else {
        cerr << "extract: syntax error, unknown reordering model: " << modelName << endl;
        exit(1);
      }

      allModelsOutputFlag = true;
    } else {
      cerr << "extract: syntax error, unknown option '" << string(argv[i]) << "'\n";
      exit(1);
    }
  }

  // default reordering model if no model selected
  // allows for the old syntax to be used
  if(orientationFlag && !allModelsOutputFlag) {
    wordModel = true;
    wordType = REO_MSD;
  }

  // open input files
  Moses::InputFileStream eFile(fileNameE);
  Moses::InputFileStream fFile(fileNameF);
  Moses::InputFileStream aFile(fileNameA);

  istream *eFileP = &eFile;
  istream *fFileP = &fFile;
  istream *aFileP = &aFile;
//.........这里部分代码省略.........