uint AddWord(
const uint sentno,
const AnchoredPhrasePair &app,
const uint maxLength
) {
WordAlignment wa = app.second.get().getWordAlignment();
PhraseData sd = app.second.get().getSourcePhrase().get();
PhraseData td = app.second.get().getTargetPhrase().get();
uint addCount = 0;
for(uint j = 0; j < sd.size(); ++j) {
// just for testing: words longer than 5 characters .... (should use some other criteria here!)
if((maxLength == 0) || (sd[j].size() > maxLength)) {
for(WordAlignment::const_iterator
wit = wa.begin_for_source(j);
wit != wa.end_for_source(j);
++wit
) {
sentWords[sentno].push_back(td[*wit]);
addCount++;
// LOG(logger_, debug, "add word " << td[*wit] << " aligned to " << sd[j]);
}
}
}
return addCount;
}
uint AddWord(
const uint sentno,
const uint phrno,
const AnchoredPhrasePair &app,
const std::string pos,
const int historySize
) {
WordAlignment wa = app.second.get().getWordAlignment();
PhraseData sd = app.second.get().getSourcePhrase().get();
PhraseData td = app.second.get().getTargetPhrase().get();
uint wordno = app.first.find_first();
uint addCount = 0;
for(uint j = 0; j < sd.size(); ++j) {
// TODO: we could support other conditins here as well!
if(posTags[sentno][wordno] == pos) {
for(WordAlignment::const_iterator
wit = wa.begin_for_source(j);
wit != wa.end_for_source(j);
++wit
) {
std::string wordPair = sd[j] + "_" + td[*wit];
long long b = FindVocabularyPosition(wordPair);
if(b < 0)
continue;
SelectedWordVector word(phrno,*wit,sd[j],td[*wit],wordno,size);
FindVector(b,word.vec);
selectedWords[sentno].push_back(word);
selectedWords[sentno].back().similarity = MaxSimilarityWithHistory(
sentno,
selectedWords[sentno].size()-1,
historySize
);
// currentScore += word.similarity;
addCount++;
//LOG(logger_, debug, "add word " << td[*wit] << " aligned to " << sd[j]);
}
}
wordno++;
}
return addCount;
};
