// the most general, but probably also most inefficient implementation
void lemur::api::RetrievalMethod::scoreCollection(const QueryRep &qry, IndexedRealVector &results)
{
results.clear();
for (DOCID_T id=1; id<=ind.docCount();id++) {
results.PushValue(id, scoreDoc(qry, id));
}
}
// keywordToDocs '+indexPath+' "'+query+'"','r'
int main(int argc, const char* argv[])
{
if (argc == 2) {
if (!strcmp(argv[1], "-h") || !strcmp(argv[1], "--help")) {
cout << endl << "Usage:" << endl << endl <<
"keywordToDocs indexPath \"query\"" << endl << endl <<
"Returns a list of doc numbers that contain the query" <<
endl << endl;
return(0);
}
}
Index *ind = IndexManager::openIndex(argv[1]);
ArrayAccumulator accumulator(ind->docCount());
RetrievalMethod *myMethod = new TFIDFRetMethod(*ind, accumulator);
IndexedRealVector results;
StringQuery *q = new StringQuery(argv[2]); // construct a TextQuery
QueryRep * qr = myMethod->computeQueryRep(*q); // compute the query representation
// now score all documents
myMethod->scoreCollection(*qr, results);
results.Sort(); // sorting results, assume a higher score means more relevant
IndexedRealVector::iterator it;
it = results.begin();
while ((it != results.end())) {
cout << (*it).ind // this is the document ID
<< endl;
it++;
}
}
/// Score a set of documents w.r.t. a query rep (e.g. for re-ranking)
void lemur::api::RetrievalMethod::scoreDocSet(const QueryRep &qry, const DocIDSet &docSet, IndexedRealVector &results)
{
results.clear();
docSet.startIteration();
while (docSet.hasMore()) {
int docID;
double prevScore;
docSet.nextIDInfo(docID, prevScore);
results.PushValue(docID, scoreDoc(qry, docID));
}
}
/** Score a set of documents w.r.t. a query rep (e.g. for re-ranking) */
void LinkDistributionRetMethod::scoreDocSet(const QueryRep &qry, const DocIDSet &docSet, IndexedRealVector &results)
{
results.clear();
docSet.startIteration();
while(docSet.hasMore())
{
int docID;
double prevScore;
docSet.nextIDInfo(docID, prevScore);
double phi = lemur::api::ParamGetDouble("phi", 0);
double currentScore = (1 - phi) * prevScore + phi * scoreDoc(qry, docID);
results.PushValue(docID, currentScore);
}//while
}//end scoreDocSet
void LinkDistributionRetMethod::scoreDocSet(const Query &q, const DocIDSet &docSet, IndexedRealVector &results)
{
//assert(LinkDistributionParameter::docWindowSize>0);
results.clear();
docSet.startIteration();
while(docSet.hasMore())
{
int docID;
double prevScore;
docSet.nextIDInfo(docID, prevScore);
//cerr<<"docID:"<<docID<<endl;
double phi = lemur::api::ParamGetDouble("phi", 0);
double currentScore = (1 - phi) * prevScore + phi * scoreDoc(q, docID);
results.PushValue(docID, currentScore);
}//while
}
void lemur::api::StructQueryRetMethod::scoreInvertedIndex(const QueryRep &qRep,
IndexedRealVector &scores,
bool scoreAll) {
COUNT_T numDocs = ind.docCount();
COUNT_T i;
lemur::retrieval::QueryNode *queryRoot = ((lemur::retrieval::StructQueryRep *)(&qRep))->topnode();
scores.clear();
// eval the query
StructQueryScoreFunc * scorer = (StructQueryScoreFunc *)scoreFunc();
for (i = 1; i <= numDocs; i++) {
if (queryRoot->dList[i]) {
DocumentRep *dRep = computeDocRep(i);
scores.PushValue(i, scorer->evalQuery(queryRoot, dRep));
delete dRep;
} else if (scoreAll) {
scores.PushValue(i, scorer->adjustedScore(0, queryRoot));
}
}
}
void Retrieval(double *qryArr, IndexedRealVector &results, Index *ind){
//retrieve documents with respect to the array representation of the query
lemur::retrieval::ArrayAccumulator scoreAccumulator(ind->docCount());
scoreAccumulator.reset();
for (int t=1; t<=ind->termCountUnique();t++) {
if (qryArr[t]>0) {
// fetch inverted entries for a specific query term
DocInfoList *docList = ind->docInfoList(t);
// iterate over all individual documents
docList->startIteration();
while (docList->hasMore()) {
DocInfo *matchInfo = docList->nextEntry();
// for each matched term, calculated the evidence
double wt;
if (strcmp(LocalParameter::weightScheme.c_str(),"RawTF")==0) {
wt = computeRawTFWeight(matchInfo->docID(), // doc ID
t, // term ID
matchInfo->termCount(), // freq of term t in this doc
qryArr[t], // freq of term t in the query
ind);
}else if (strcmp(LocalParameter::weightScheme.c_str(),"RawTFIDF")==0) {
wt = computeRawTFIDFWeight(matchInfo->docID(), // doc ID
t, // term ID
matchInfo->termCount(), // freq of term t in this doc
qryArr[t], // freq of term t in the query
ind);
}else if (strcmp(LocalParameter::weightScheme.c_str(),"LogTFIDF")==0) {
wt = computeLogTFIDFWeight(matchInfo->docID(), // doc ID
t, // term ID
matchInfo->termCount(), // freq of term t in this doc
qryArr[t], // freq of term t in the query
ind);
}else if (strcmp(LocalParameter::weightScheme.c_str(),"Okapi")==0) {
wt = computeOkapiWeight(matchInfo->docID(), // doc ID
t, // term ID
matchInfo->termCount(), // freq of term t in this doc
qryArr[t], // freq of term t in the query
ind);
}else if (strcmp(LocalParameter::weightScheme.c_str(),"Custom")==0){
wt = computeCustomWeight(matchInfo->docID(), // doc ID
t, // term ID
matchInfo->termCount(), // freq of term t in this doc
qryArr[t], // freq of term t in the query
ind);
}else{
cerr<<"The weighting scheme of "<<LocalParameter::weightScheme.c_str()<<" is not supported"<<endl;
exit(1);
}
scoreAccumulator.incScore(matchInfo->docID(),wt);
}
delete docList;
}
}
// Adjust the scores for the documents when it is necessary
double s;
int d;
for (d=1; d<=ind->docCount(); d++) {
if (scoreAccumulator.findScore(d,s)) {
} else {
s=0;
}
if (strcmp(LocalParameter::weightScheme.c_str(),"RawTF")==0) {
results.PushValue(d, computeAdjustedScore(s, // the score from the accumulator
d, // doc ID
ind)); // index
}else if (strcmp(LocalParameter::weightScheme.c_str(),"RawTFIDF")==0) {
results.PushValue(d, computeAdjustedScore(s, // the score from the accumulator
d, // doc ID
ind)); // index
}else if (strcmp(LocalParameter::weightScheme.c_str(),"LogTFIDF")==0) {
results.PushValue(d, computeAdjustedScore(s, // the score from the accumulator
d, // doc ID
ind)); // index
}else if (strcmp(LocalParameter::weightScheme.c_str(),"Okapi")==0) {
results.PushValue(d, computeAdjustedScore(s, // the score from the accumulator
d, // doc ID
ind)); // index
}else if (strcmp(LocalParameter::weightScheme.c_str(),"Custom")==0){
results.PushValue(d, computeCustomAdjustedScore(s, // the score from the accumulator
d, // doc ID
ind)); // index
}else{
cerr<<"The weighting scheme of "<<LocalParameter::weightScheme.c_str()<<" is not supported"<<endl;
exit(1);
}
}
}
