void interpolateWith(const lemur::langmod::UnigramLM &qModel,
double origModCoeff,
int howManyWord,
double prSumThresh,
double prThresh) {
if (!qm) {
qm = new lemur::api::IndexedRealVector();
} else {
qm->clear();
}
qModel.startIteration();
while (qModel.hasMore()) {
IndexedReal entry;
qModel.nextWordProb((TERMID_T &)entry.ind,entry.val);
qm->push_back(entry);
}
qm->Sort();
double countSum = totalCount();
startIteration();
while (hasMore()) {
QueryTerm *qt = nextTerm();
setCount(qt->id(), qt->weight()*origModCoeff/countSum);
delete qt;
}
cout << "-------- FB terms --------" << endl;
double prSum = 0;
int wdCount = 0;
IndexedRealVector::iterator it;
it = qm->begin();
while (it != qm->end() && prSum < prSumThresh &&
wdCount < howManyWord && (*it).val >=prThresh) {
incCount((*it).ind, (*it).val*(1-origModCoeff));
cout << ind.term(it->ind) << endl;
prSum += (*it).val;
it++;
wdCount++;
}
cout << "--------------------------" << endl;
colQLikelihood = 0;
colQueryLikelihood();
colKLComputed = false;
}
void lemur::retrieval::QueryModel::interpolateWith(const lemur::langmod::UnigramLM &qModel,
double origModCoeff,
int howManyWord,
double prSumThresh,
double prThresh) {
if (!qm) {
qm = new lemur::api::IndexedRealVector();
} else {
qm->clear();
}
qModel.startIteration();
while (qModel.hasMore()) {
IndexedReal entry;
qModel.nextWordProb((TERMID_T &)entry.ind,entry.val);
qm->push_back(entry);
}
qm->Sort();
double countSum = totalCount();
// discounting the original model
startIteration();
while (hasMore()) {
QueryTerm *qt = nextTerm();
incCount(qt->id(), qt->weight()*origModCoeff/countSum);
delete qt;
}
// now adding the new model
double prSum = 0;
int wdCount = 0;
IndexedRealVector::iterator it;
it = qm->begin();
while (it != qm->end() && prSum < prSumThresh &&
wdCount < howManyWord && (*it).val >=prThresh) {
incCount((*it).ind, (*it).val*(1-origModCoeff));
prSum += (*it).val;
it++;
wdCount++;
}
//Sum w in Q qtf * log(qtcf/termcount);
colQLikelihood = 0;
colQueryLikelihood();
colKLComputed = false;
}