static void printClarity( const std::string& query,
indri::api::QueryEnvironment & env,
const std::vector<indri::query::RelevanceModel::Gram*>& grams, int numTerms ) {
int count = 0;
double sum=0, ln_Pr=0;
for( size_t j=0; j< numTerms && j < grams.size(); j++ ) {
std::string t = grams[j]->terms[0];
count++;
// query-clarity = SUM_w{P(w|Q)*log(P(w|Q)/P(w))}
// P(w)=cf(w)/|C|
// the relevance model uses stemmed terms, so use stemCount
double pw = ((double)env.stemCount(t)/(double)env.termCount());
// P(w|Q) is a prob computed by any model, e.g. relevance models
double pwq = grams[j]->weight;
sum += pwq;
ln_Pr += (pwq)*log(pwq/pw);
}
std::cout << "# query: " << query << " = " << count << " "
<< (ln_Pr/(sum ? sum : 1.0)/log(2.0)) << std::endl;
for( size_t j=0; j< numTerms && j < grams.size(); j++ ) {
std::string t = grams[j]->terms[0];
double pw = ((double)env.stemCount(t)/(double)env.termCount());
std::cout << t << " "
<< (grams[j]->weight*log(grams[j]->weight/
// the relevance model uses stemmed terms, so use stemCount
((double)env.stemCount(t)/
(double)env.termCount())))/log(2.0) << std::endl;
}
}
SEXP collFreq(string _term) {
long res = environment.termCount(_term);
return Rcpp::wrap(res);
}
SEXP getTermCount() {
long res = environment.termCount();
return Rcpp::wrap(res);
}
