[Debtags-commits] [svn] r1815 - in debtags/1.6.0: . tools

[Enrico Zini]

Author: enrico
Date: Sat Jul  8 09:37:11 2006
New Revision: 1815

Modified:
   debtags/1.6.0/   (props changed)
   debtags/1.6.0/tools/debtags.cc
Log:
 r2998 at viaza:  enrico | 2006-07-08 11:36:17 +0200
 Improved keyword->tag mapping by weighting by cardinality when computing jumps.  Now it rocks!


Modified: debtags/1.6.0/tools/debtags.cc
==============================================================================
--- debtags/1.6.0/tools/debtags.cc	(original)
+++ debtags/1.6.0/tools/debtags.cc	Sat Jul  8 09:37:11 2006
@@ -791,27 +791,25 @@
 		return res;
 	}
 
-	TagMetrics<Tag, Number> jumpsFrom(const TagMetrics<Tag, Number>& other) const
+	TagMetrics<Tag, double> jumpsFrom(const TagMetrics<Tag, Number>& other) const
 	{
 		// Compute rank metrics
 		TagMetrics<Tag, Number> rank1 = other.rankMetrics();
 		TagMetrics<Tag, Number> rank2 = rankMetrics();
 
-		TagMetrics<Tag, Number> res;
+		TagMetrics<Tag, double> res;
 		typename TagMetrics::const_iterator i1 = rank1.begin();
 		typename TagMetrics::const_iterator i2 = rank2.begin();
 		while (i1 != rank1.end() || i2 != rank2.end())
 		{
 			if (i1->first == i2->first)
 			{
-				res.add(i1->first, i1->second - i2->second);
+				res.add(i1->first, (i1->second - i2->second) / i2->second);
 				++i1;
 				++i2;
 			} else if (i1 == rank1.end() || (i2 != rank2.end() && i2->first < i1->first)) {
-				res.add(i2->first, rank2.size() - i2->second);
 				++i2;
 			} else {
-				res.add(i1->first, i1->second - rank1.size());
 				++i1;
 			}
 		}
@@ -1111,7 +1109,7 @@
 		// Compute the set of tags that better represent the keyword search
 		TagMetrics<Tag, int> metrics1 = TagMetrics<Tag, int>::computeFromTags(fullColl);
 		collMetrics = TagMetrics<Tag, int>::computeFromTags(coll);
-		TagMetrics<Tag, int> jumps = collMetrics.jumpsFrom(metrics1);
+		TagMetrics<Tag, double> jumps = collMetrics.jumpsFrom(metrics1);
 		interesting = jumps.tagsSortedByMetrics();
 		coll = fullColl;