Changeset 36864 for gs3-extensions

Timestamp:

2022-10-24T08:33:26+13:00 (18 months ago)

Author:

davidb

Message:

Attempt at a result set that merges later matches from the same doc-id

Location:

gs3-extensions/mars-src/trunk/src/java/org/greenstone/gsdl3/util

Files:

• WekaDBDocInfo.java (modified) (1 diff)
• WekaDBWrapper.java (modified) (14 diffs)

gs3-extensions/mars-src/trunk/src/java/org/greenstone/gsdl3/util/WekaDBDocInfo.java

@@ -61 +61 @@
     }
 
+    public void incTopRank(double inc_rank_val)
+    {
+    double top_rank = rankVector_.get(0);
+
+    double new_top_rank = top_rank + inc_rank_val;
+    rankVector_.set(0,new_top_rank);
+    }
+
 
     public String getOffsetList() 

gs3-extensions/mars-src/trunk/src/java/org/greenstone/gsdl3/util/WekaDBWrapper.java

@@ -44 +44 @@
     
     /** the query result, filled in by runQuery */
-    protected Vector query_result_;
+    protected Vector query_results_;
 
     protected int offset_ = 100;
@@ -60 +60 @@
 
     public WekaDBWrapper() {  
-    query_result_ = null;
+    query_results_ = null;
     }
 
@@ -104 +104 @@
     if (first_entry) {
         WekaDBDocInfo wekaDB_doc_info = new WekaDBDocInfo(doc_id,rankVector,offsetVector);
-        query_result_.add(wekaDB_doc_info);
+        query_results_.add(wekaDB_doc_info);
         first_entry = false;
     }
@@ -112 +112 @@
         WekaDBDocInfo wekaDB_doc_info = new WekaDBDocInfo(doc_id,rank,offset);
         
-        query_result_.add(wekaDB_doc_info);
+        query_results_.add(wekaDB_doc_info);
     }
 
@@ -119 +119 @@
 
 
+    protected int mergeResultDoc(Vector query_results, WekaDBDocInfo new_doc_info, double inc_rank_val)
+    {
+    int merged = 0;
+
+    String new_doc_id = new_doc_info.getDocID();
+
+    final int query_results_len = query_results.size();
+
+    for (int i=0; i<query_results_len; i++) {
+        WekaDBDocInfo existing_doc_info = (WekaDBDocInfo)query_results.get(i);
+
+        String existing_doc_id = existing_doc_info.getDocID();
+        if (new_doc_id.equals(existing_doc_id)) {
+        merged = 1;
+        existing_doc_info.incTopRank(inc_rank_val);
+        break;
+        }
+    }
+
+    if (merged == 0) {
+        query_results.add(new_doc_info);
+    }
+    
+    return merged;
+    }
+    
     /** actually carry out the query.
     Use the set methods to set query results.
-    Writes the result to query_result.
+    Writes the result to query_results.
      * - maintains state between requests as can be slow  
      * base_dir and index_path should join together to provide
@@ -172 +198 @@
 
         
-    Vector expanded_query_result = new Vector();
+    Vector expanded_query_results = new Vector();
 
     int nearest_instances_len = nearest_instances.size();
@@ -179 +205 @@
 
     double pos_penalty = 0.1;
+    int    topup_count = 0;
     
     for (int ei=0; ei<clamped_expanded_k_nearest_num; ei++) {
@@ -186 +213 @@
         String matching_doc_id_segment = instance.stringValue(0);
 
-        
         //Pattern p = Pattern.compile("^(\\w+)-(\\d+)$");
         Matcher m = doc_seg_re.matcher(matching_doc_id_segment);
@@ -193 +219 @@
         String matching_doc_id = m.group(1);
         int end_of_matching_segment_offset = Integer.parseInt(m.group(2));
-        int matching_segment_offset = end_of_matching_segment_offset - (int)AV_SEGMENT_LENGTH_SECS;
+        //int matching_segment_offset = end_of_matching_segment_offset - (int)AV_SEGMENT_LENGTH_SECS;
+        int matching_segment_offset = end_of_matching_segment_offset;
         
         if (matching_doc_id.equals(doc_id)) {
+            // don't add in matches that come from a matching segment in the query doc
             continue;
         }
@@ -208 +236 @@
         logger.info("\tAdding in: matching_doc_id = " + matching_doc_id);
         WekaDBDocInfo wekaDB_doc_info = new WekaDBDocInfo(matching_doc_id,matching_rank,matching_segment_offset);
-        expanded_query_result.add(wekaDB_doc_info);
+        
+        //expanded_query_results.add(wekaDB_doc_info);
+
+        double inc_rank_val = matching_rank / (double)(topup_count+2); // starts to a 50% (/2) weighting when topup_count == 0
+        int merged = mergeResultDoc(expanded_query_results,wekaDB_doc_info,inc_rank_val);
+
+        topup_count += merged;
+
+        if ((expanded_query_results.size() > k_nearest_num) && (topup_count > k_nearest_num)) {
+            // guard to stop multiple recurring matches in the same doc dominationg the rank_val
+            break;
+        }
         }
         else {
@@ -214 +253 @@
         }          
     }
-    
-    query_result_ = new Vector();
+
+    Collections.sort(expanded_query_results);
+        
+    query_results_ = new Vector();
 
     int i = 0;
     while (i < k_nearest_num) {
-        if (i >= expanded_query_result.size()) {
+        if (i >= expanded_query_results.size()) {
         break;
         }
         
-        query_result_.add(expanded_query_result.get(i));
+        query_results_.add(expanded_query_results.get(i));
         i++;
     }
 
-    Collections.sort(query_result_);    
+    //Collections.sort(query_results_); 
     }
     
@@ -265 +306 @@
         BufferedReader wbr = new BufferedReader(wisr);
 
-        query_result_ = new Vector();
+        query_results_ = new Vector();
 
         boolean first_entry = true;
@@ -335 +376 @@
         wbr.close();
 
-        // sort query_result_ on 'rank' field 
+        // sort query_results_ on 'rank' field 
         // note: compareTo() method impelemented to sort into descending order
 
-        Collections.sort(query_result_);
+        Collections.sort(query_results_);
 
 
@@ -353 +394 @@
     public Vector getQueryResult()
     {
-    return query_result_;
+    return query_results_;
     }
 }