void ExtractTask::saveAllHieroPhrases( int startT, int endT, int startS, int endS, HoleCollection &holeColl, int countS)
{
LabelIndex labelIndex,labelCount;
// number of target head labels
int numLabels = m_options.targetSyntax ? m_sentence.targetTree.GetNodes(startT,endT).size() : 1;
if (m_options.targetSyntacticPreferences && !numLabels) {
numLabels++;
}
labelCount.push_back(numLabels);
labelIndex.push_back(0);
// number of source head labels
numLabels = m_options.sourceSyntax ? m_sentence.sourceTree.GetNodes(startS,endS).size() : 1;
labelCount.push_back(numLabels);
labelIndex.push_back(0);
// number of target hole labels
for( HoleList::const_iterator hole = holeColl.GetHoles().begin();
hole != holeColl.GetHoles().end(); hole++ ) {
int numLabels = m_options.targetSyntax ? m_sentence.targetTree.GetNodes(hole->GetStart(1),hole->GetEnd(1)).size() : 1 ;
if (m_options.targetSyntacticPreferences && !numLabels) {
numLabels++;
}
labelCount.push_back(numLabels);
labelIndex.push_back(0);
}
// number of source hole labels
holeColl.SortSourceHoles();
for( vector<Hole*>::iterator i = holeColl.GetSortedSourceHoles().begin();
i != holeColl.GetSortedSourceHoles().end(); i++ ) {
const Hole &hole = **i;
int numLabels = m_options.sourceSyntax ? m_sentence.sourceTree.GetNodes(hole.GetStart(0),hole.GetEnd(0)).size() : 1 ;
labelCount.push_back(numLabels);
labelIndex.push_back(0);
}
// loop through the holes
bool done = false;
while(!done) {
saveHieroPhrase( startT, endT, startS, endS, holeColl, labelIndex, countS );
for(unsigned int i=0; i<labelIndex.size(); i++) {
labelIndex[i]++;
if(labelIndex[i] == labelCount[i]) {
labelIndex[i] = 0;
if (i == labelIndex.size()-1)
done = true;
} else {
break;
}
}
}
}
void printAllHieroPhrases( SentenceAlignmentWithSyntax &sentence
, int startT, int endT, int startS, int endS
, HoleCollection &holeColl)
{
LabelIndex labelIndex,labelCount;
// number of target head labels
int numLabels = options.targetSyntax ? sentence.targetTree.GetNodes(startT,endT).size() : 1;
labelCount.push_back(numLabels);
labelIndex.push_back(0);
// number of source head labels
numLabels = options.sourceSyntax ? sentence.sourceTree.GetNodes(startS,endS).size() : 1;
labelCount.push_back(numLabels);
labelIndex.push_back(0);
// number of target hole labels
for( HoleList::const_iterator hole = holeColl.GetHoles().begin();
hole != holeColl.GetHoles().end(); hole++ ) {
int numLabels = options.targetSyntax ? sentence.targetTree.GetNodes(hole->GetStart(1),hole->GetEnd(1)).size() : 1 ;
labelCount.push_back(numLabels);
labelIndex.push_back(0);
}
// number of source hole labels
holeColl.SortSourceHoles();
for( vector<Hole*>::iterator i = holeColl.GetSortedSourceHoles().begin();
i != holeColl.GetSortedSourceHoles().end(); i++ ) {
const Hole &hole = **i;
int numLabels = options.sourceSyntax ? sentence.sourceTree.GetNodes(hole.GetStart(0),hole.GetEnd(0)).size() : 1 ;
labelCount.push_back(numLabels);
labelIndex.push_back(0);
}
// loop through the holes
bool done = false;
while(!done) {
printHieroPhrase( sentence, startT, endT, startS, endS, holeColl, labelIndex );
for(int i=0; i<labelIndex.size(); i++) {
labelIndex[i]++;
if(labelIndex[i] == labelCount[i]) {
labelIndex[i] = 0;
if (i == labelIndex.size()-1)
done = true;
} else {
break;
}
}
}
}
string printSourceHieroPhrase( SentenceAlignmentWithSyntax &sentence
, int startT, int endT, int startS, int endS
, HoleCollection &holeColl, const LabelIndex &labelIndex)
{
vector<Hole*>::iterator iterHoleList = holeColl.GetSortedSourceHoles().begin();
assert(iterHoleList != holeColl.GetSortedSourceHoles().end());
bool stringToTree = !options.sourceSyntax && options.targetSyntax;
string out = "";
int outPos = 0;
int holeCount = 0;
for(int currPos = startS; currPos <= endS; currPos++) {
bool isHole = false;
if (iterHoleList != holeColl.GetSortedSourceHoles().end()) {
const Hole &hole = **iterHoleList;
isHole = hole.GetStart(0) == currPos;
}
if (isHole) {
Hole &hole = **iterHoleList;
const string &targetLabel = hole.GetLabel(1);
assert(targetLabel != "");
const string &sourceLabel = hole.GetLabel(0);
if (stringToTree) {
out += "[" + sourceLabel + "] ";
} else {
out += "[" + sourceLabel + "][" + targetLabel + "] ";
}
currPos = hole.GetEnd(0);
hole.SetPos(outPos, 0);
++iterHoleList;
++holeCount;
} else {
out += sentence.source[currPos] + " ";
}
outPos++;
}
assert(iterHoleList == holeColl.GetSortedSourceHoles().end());
return out.erase(out.size()-1);
}
string ExtractTask::saveSourceHieroPhrase( int startT, int endT, int startS, int endS
, HoleCollection &holeColl, const LabelIndex &labelIndex)
{
vector<Hole*>::iterator iterHoleList = holeColl.GetSortedSourceHoles().begin();
assert(iterHoleList != holeColl.GetSortedSourceHoles().end());
string out = "";
int outPos = 0;
int holeCount = 0;
for(int currPos = startS; currPos <= endS; currPos++) {
bool isHole = false;
if (iterHoleList != holeColl.GetSortedSourceHoles().end()) {
const Hole &hole = **iterHoleList;
isHole = hole.GetStart(0) == currPos;
}
if (isHole) {
Hole &hole = **iterHoleList;
const string &targetLabel = hole.GetLabel(1);
assert(targetLabel != "");
const string &sourceLabel = hole.GetLabel(0);
if (m_options.unpairedExtractFormat) {
out += "[" + sourceLabel + "] ";
} else {
out += "[" + sourceLabel + "][" + targetLabel + "] ";
}
currPos = hole.GetEnd(0);
hole.SetPos(outPos, 0);
++iterHoleList;
++holeCount;
} else {
out += m_sentence.source[currPos] + " ";
}
outPos++;
}
assert(iterHoleList == holeColl.GetSortedSourceHoles().end());
return out.erase(out.size()-1);
}
void preprocessSourceHieroPhrase( SentenceAlignmentWithSyntax &sentence
, int startT, int endT, int startS, int endS
, WordIndex &indexS, HoleCollection &holeColl, const LabelIndex &labelIndex)
{
vector<Hole*>::iterator iterHoleList = holeColl.GetSortedSourceHoles().begin();
assert(iterHoleList != holeColl.GetSortedSourceHoles().end());
int outPos = 0;
int holeCount = 0;
int holeTotal = holeColl.GetHoles().size();
for(int currPos = startS; currPos <= endS; currPos++) {
bool isHole = false;
if (iterHoleList != holeColl.GetSortedSourceHoles().end()) {
const Hole &hole = **iterHoleList;
isHole = hole.GetStart(0) == currPos;
}
if (isHole) {
Hole &hole = **iterHoleList;
int labelI = labelIndex[ 2+holeCount+holeTotal ];
string label = options.sourceSyntax ?
sentence.sourceTree.GetNodes(currPos,hole.GetEnd(0))[ labelI ]->GetLabel() : "X";
hole.SetLabel(label, 0);
currPos = hole.GetEnd(0);
hole.SetPos(outPos, 0);
++iterHoleList;
++holeCount;
} else {
indexS[currPos] = outPos;
}
outPos++;
}
assert(iterHoleList == holeColl.GetSortedSourceHoles().end());
}