Merge pull request meta-toolkit#89 from husseinhazimeh/master

Change query_term_count -> query_term_weight to support non-integer weighted queries.
henfiber · Jun 3, 2015 · a5f7688 · a5f7688
2 parents 39752d5 + e2ead62
commit a5f7688
Show file tree

Hide file tree

Showing 5 changed files with 7 additions and 7 deletions.
diff --git a/include/index/score_data.h b/include/index/score_data.h
@@ -55,8 +55,8 @@ struct score_data
 
     /// doc term id
     term_id t_id;
-    /// query term count
-    uint64_t query_term_count;
+    /// query term count (or weight in case of feedback)
+    double query_term_weight;
     /// number of docs that t_id appears in
     uint64_t doc_count;
     /// number of times t_id appears in corpus

diff --git a/src/index/ranker/lm_ranker.cpp b/src/index/ranker/lm_ranker.cpp
@@ -20,7 +20,7 @@ double language_model_ranker::score_one(const score_data& sd)
     double ps = smoothed_prob(sd);
     double pc = static_cast<double>(sd.corpus_term_count) / sd.total_terms;
 
-    return sd.query_term_count * std::log(ps / (doc_constant(sd) * pc));
+    return sd.query_term_weight * std::log(ps / (doc_constant(sd) * pc));
 }
 
 double language_model_ranker::initial_score(const score_data& sd) const

diff --git a/src/index/ranker/okapi_bm25.cpp b/src/index/ranker/okapi_bm25.cpp
@@ -32,8 +32,8 @@ double okapi_bm25::score_one(const score_data& sd)
                 / ((k1_ * ((1.0 - b_) + b_ * doc_len / sd.avg_dl))
                    + sd.doc_term_count);
 
-    double QTF = ((k3_ + 1.0) * sd.query_term_count)
-                 / (k3_ + sd.query_term_count);
+    double QTF = ((k3_ + 1.0) * sd.query_term_weight)
+                 / (k3_ + sd.query_term_weight);
 
     return TF * IDF * QTF;
 }

diff --git a/src/index/ranker/pivoted_length.cpp b/src/index/ranker/pivoted_length.cpp
@@ -26,7 +26,7 @@ double pivoted_length::score_one(const score_data& sd)
     double norm = (1 - s_) + s_ * (doc_len / sd.avg_dl);
     double IDF = log((sd.num_docs + 1) / (0.5 + sd.doc_count));
 
-    return TF / norm * sd.query_term_count * IDF;
+    return TF / norm * sd.query_term_weight * IDF;
 }
 
 template <>

diff --git a/src/index/ranker/ranker.cpp b/src/index/ranker/ranker.cpp
@@ -36,7 +36,7 @@ ranker::score(inverted_index& idx, corpus::document& query,
         auto pdata = idx.search_primary(t_id);
         sd.doc_count = pdata->counts().size();
         sd.t_id = t_id;
-        sd.query_term_count = tpair.second;
+        sd.query_term_weight = tpair.second;
         sd.corpus_term_count = idx.total_num_occurences(sd.t_id);
         for (auto& dpair : pdata->counts())
         {