C++ Hypothesis::GetInput方法代码示例

float
DistortionScoreProducer::
CalculateDistortionScore(const Hypothesis& hypo,
                         const Range &prev, const Range &curr, const int FirstGap)
{
  // if(!StaticData::Instance().UseEarlyDistortionCost()) {
  if(!hypo.GetManager().options()->reordering.use_early_distortion_cost) {
    return - (float) hypo.GetInput().ComputeDistortionDistance(prev, curr);
  } // else {

  /* Pay distortion score as soon as possible, from Moore and Quirk MT Summit 2007
     Definitions:
     S   : current source range
     S'  : last translated source phrase range
     S'' : longest fully-translated initial segment
  */

  int prefixEndPos = (int)FirstGap-1;
  if((int)FirstGap==-1)
    prefixEndPos = -1;

  // case1: S is adjacent to S'' => return 0
  if ((int) curr.GetStartPos() == prefixEndPos+1) {
    IFVERBOSE(4) std::cerr<< "MQ07disto:case1" << std::endl;
    return 0;
  }

  // case2: S is to the left of S' => return 2(length(S))
  if ((int) curr.GetEndPos() < (int) prev.GetEndPos()) {
    IFVERBOSE(4) std::cerr<< "MQ07disto:case2" << std::endl;
    return (float) -2*(int)curr.GetNumWordsCovered();
  }

  // case3: S' is a subsequence of S'' => return 2(nbWordBetween(S,S'')+length(S))
  if ((int) prev.GetEndPos() <= prefixEndPos) {
    IFVERBOSE(4) std::cerr<< "MQ07disto:case3" << std::endl;
    int z = (int)curr.GetStartPos()-prefixEndPos - 1;
    return (float) -2*(z + (int)curr.GetNumWordsCovered());
  }

  // case4: otherwise => return 2(nbWordBetween(S,S')+length(S))
  IFVERBOSE(4) std::cerr<< "MQ07disto:case4" << std::endl;
  return (float) -2*((int)curr.GetNumWordsBetween(prev) + (int)curr.GetNumWordsCovered());

}

void WordTranslationFeature::EvaluateWhenApplied
(const Hypothesis& hypo,
 ScoreComponentCollection* accumulator) const
{
  const Sentence& input = static_cast<const Sentence&>(hypo.GetInput());
  const TranslationOption& transOpt = hypo.GetTranslationOption();
  const TargetPhrase& targetPhrase = hypo.GetCurrTargetPhrase();
  const AlignmentInfo &alignment = targetPhrase.GetAlignTerm();

  // process aligned words
  for (AlignmentInfo::const_iterator alignmentPoint = alignment.begin(); alignmentPoint != alignment.end(); alignmentPoint++) {
    const Phrase& sourcePhrase = transOpt.GetInputPath().GetPhrase();
    int sourceIndex = alignmentPoint->first;
    int targetIndex = alignmentPoint->second;
    Word ws = sourcePhrase.GetWord(sourceIndex);
    if (m_factorTypeSource == 0 && ws.IsNonTerminal()) continue;
    Word wt = targetPhrase.GetWord(targetIndex);
    if (m_factorTypeSource == 0 && wt.IsNonTerminal()) continue;
    StringPiece sourceWord = ws.GetFactor(m_factorTypeSource)->GetString();
    StringPiece targetWord = wt.GetFactor(m_factorTypeTarget)->GetString();
    if (m_ignorePunctuation) {
      // check if source or target are punctuation
      char firstChar = sourceWord[0];
      CharHash::const_iterator charIterator = m_punctuationHash.find( firstChar );
      if(charIterator != m_punctuationHash.end())
        continue;
      firstChar = targetWord[0];
      charIterator = m_punctuationHash.find( firstChar );
      if(charIterator != m_punctuationHash.end())
        continue;
    }

    if (!m_unrestricted) {
      if (FindStringPiece(m_vocabSource, sourceWord) == m_vocabSource.end())
        sourceWord = "OTHER";
      if (FindStringPiece(m_vocabTarget, targetWord) == m_vocabTarget.end())
        targetWord = "OTHER";
    }

    if (m_simple) {
      // construct feature name
      stringstream featureName;
      featureName << m_description << "_";
      featureName << sourceWord;
      featureName << "~";
      featureName << targetWord;
      accumulator->SparsePlusEquals(featureName.str(), 1);
    }
    if (m_domainTrigger && !m_sourceContext) {
      const bool use_topicid = input.GetUseTopicId();
      const bool use_topicid_prob = input.GetUseTopicIdAndProb();
      if (use_topicid || use_topicid_prob) {
        if(use_topicid) {
          // use topicid as trigger
          const long topicid = input.GetTopicId();
          stringstream feature;
          feature << m_description << "_";
          if (topicid == -1)
            feature << "unk";
          else
            feature << topicid;

          feature << "_";
          feature << sourceWord;
          feature << "~";
          feature << targetWord;
          accumulator->SparsePlusEquals(feature.str(), 1);
        } else {
          // use topic probabilities
          const vector<string> &topicid_prob = *(input.GetTopicIdAndProb());
          if (atol(topicid_prob[0].c_str()) == -1) {
            stringstream feature;
            feature << m_description << "_unk_";
            feature << sourceWord;
            feature << "~";
            feature << targetWord;
            accumulator->SparsePlusEquals(feature.str(), 1);
          } else {
            for (size_t i=0; i+1 < topicid_prob.size(); i+=2) {
              stringstream feature;
              feature << m_description << "_";
              feature << topicid_prob[i];
              feature << "_";
              feature << sourceWord;
              feature << "~";
              feature << targetWord;
              accumulator->SparsePlusEquals(feature.str(), atof((topicid_prob[i+1]).c_str()));
            }
          }
        }
      } else {
        // range over domain trigger words (keywords)
        const long docid = input.GetDocumentId();
        for (boost::unordered_set<std::string>::const_iterator p = m_vocabDomain[docid].begin(); p != m_vocabDomain[docid].end(); ++p) {
          string sourceTrigger = *p;
          stringstream feature;
          feature << m_description << "_";
          feature << sourceTrigger;
          feature << "_";
          feature << sourceWord;
//.........这里部分代码省略.........

void PhrasePairFeature::EvaluateWhenApplied(
  const Hypothesis& hypo,
  ScoreComponentCollection* accumulator) const
{
  const TargetPhrase& target = hypo.GetCurrTargetPhrase();
  const Phrase& source = hypo.GetTranslationOption().GetInputPath().GetPhrase();
  if (m_simple) {
    ostringstream namestr;
    namestr << "pp_";
    namestr << source.GetWord(0).GetFactor(m_sourceFactorId)->GetString();
    for (size_t i = 1; i < source.GetSize(); ++i) {
      const Factor* sourceFactor = source.GetWord(i).GetFactor(m_sourceFactorId);
      namestr << ",";
      namestr << sourceFactor->GetString();
    }
    namestr << "~";
    namestr << target.GetWord(0).GetFactor(m_targetFactorId)->GetString();
    for (size_t i = 1; i < target.GetSize(); ++i) {
      const Factor* targetFactor = target.GetWord(i).GetFactor(m_targetFactorId);
      namestr << ",";
      namestr << targetFactor->GetString();
    }

    accumulator->SparsePlusEquals(namestr.str(),1);
  }
  if (m_domainTrigger) {
    const Sentence& input = static_cast<const Sentence&>(hypo.GetInput());
    const bool use_topicid = input.GetUseTopicId();
    const bool use_topicid_prob = input.GetUseTopicIdAndProb();

    // compute pair
    ostringstream pair;
    pair << source.GetWord(0).GetFactor(m_sourceFactorId)->GetString();
    for (size_t i = 1; i < source.GetSize(); ++i) {
      const Factor* sourceFactor = source.GetWord(i).GetFactor(m_sourceFactorId);
      pair << ",";
      pair << sourceFactor->GetString();
    }
    pair << "~";
    pair << target.GetWord(0).GetFactor(m_targetFactorId)->GetString();
    for (size_t i = 1; i < target.GetSize(); ++i) {
      const Factor* targetFactor = target.GetWord(i).GetFactor(m_targetFactorId);
      pair << ",";
      pair << targetFactor->GetString();
    }

    if (use_topicid || use_topicid_prob) {
      if(use_topicid) {
        // use topicid as trigger
        const long topicid = input.GetTopicId();
        stringstream feature;
        feature << "pp_";
        if (topicid == -1)
          feature << "unk";
        else
          feature << topicid;

        feature << "_";
        feature << pair.str();
        accumulator->SparsePlusEquals(feature.str(), 1);
      } else {
        // use topic probabilities
        const vector<string> &topicid_prob = *(input.GetTopicIdAndProb());
        if (atol(topicid_prob[0].c_str()) == -1) {
          stringstream feature;
          feature << "pp_unk_";
          feature << pair.str();
          accumulator->SparsePlusEquals(feature.str(), 1);
        } else {
          for (size_t i=0; i+1 < topicid_prob.size(); i+=2) {
            stringstream feature;
            feature << "pp_";
            feature << topicid_prob[i];
            feature << "_";
            feature << pair.str();
            accumulator->SparsePlusEquals(feature.str(), atof((topicid_prob[i+1]).c_str()));
          }
        }
      }
    } else {
      // range over domain trigger words
      const long docid = input.GetDocumentId();
      for (set<string>::const_iterator p = m_vocabDomain[docid].begin(); p != m_vocabDomain[docid].end(); ++p) {
        string sourceTrigger = *p;
        ostringstream namestr;
        namestr << "pp_";
        namestr << sourceTrigger;
        namestr << "_";
        namestr << pair.str();
        accumulator->SparsePlusEquals(namestr.str(),1);
      }
    }
  }
  if (m_sourceContext) {
    const Sentence& input = static_cast<const Sentence&>(hypo.GetInput());

    // range over source words to get context
    for(size_t contextIndex = 0; contextIndex < input.GetSize(); contextIndex++ ) {
      StringPiece sourceTrigger = input.GetWord(contextIndex).GetFactor(m_sourceFactorId)->GetString();
      if (m_ignorePunctuation) {
//.........这里部分代码省略.........