Changeset 12775

Timestamp:

2006-09-19T12:55:21+12:00 (18 years ago)

Author:

mdewsnip

Message:

Fixed bug where some terms have zero frequency (because they don't actually appear in the matching documents).

Location:

trunk

Files:

• gsdl/bin/java/LuceneWrap.jar (modified) ( previous)
• gsdl/src/java/org/nzdl/gsdl/LuceneWrap/GS2LuceneQuery.java (modified) (1 diff)
• indexers/lucene-gs/src/org/greenstone/LuceneWrapper/GS2LuceneQuery.java (modified) (1 diff)

trunk/gsdl/src/java/org/nzdl/gsdl/LuceneWrap/GS2LuceneQuery.java

@@ -215 +215 @@
             Integer count = (Integer) term_counts.get(term);
             String field = (String) term_fields.get(term);
-            System.out.println("  <Term value=\"" + term + "\" field=\"" + field + "\" freq=\"" + count.intValue() + "\" />");
+
+            // Ignore any terms with zero frequency, because they don't exist in the matching
+            // documents. It seems that this should never happen, but it's a consequence of
+            // how the terms are identified. The terms are found by rewriting the query (above).
+            // At this point, the query hasn't been run, so each query term is expanded without
+            // knowing whether the expanded term will actually appear in one of the resulting
+            // documents. For example, "otago" may be expanded to "otaio" and "otaqo", but if
+            // the search is for "otago AND auckland", no matching documents may include "otaio".
+            // Hopefully that made some sense...
+            if (count.intValue() > 0) {
+                System.out.println("  <Term value=\"" + term + "\" field=\"" + field + "\" freq=\"" + count.intValue() + "\" />");
+            }
             count = null;
             term = null;
             }
+
             // Cleanup
             terms_iter = null;

trunk/indexers/lucene-gs/src/org/greenstone/LuceneWrapper/GS2LuceneQuery.java

@@ -215 +215 @@
             Integer count = (Integer) term_counts.get(term);
             String field = (String) term_fields.get(term);
-            System.out.println("  <Term value=\"" + term + "\" field=\"" + field + "\" freq=\"" + count.intValue() + "\" />");
+
+            // Ignore any terms with zero frequency, because they don't exist in the matching
+            // documents. It seems that this should never happen, but it's a consequence of
+            // how the terms are identified. The terms are found by rewriting the query (above).
+            // At this point, the query hasn't been run, so each query term is expanded without
+            // knowing whether the expanded term will actually appear in one of the resulting
+            // documents. For example, "otago" may be expanded to "otaio" and "otaqo", but if
+            // the search is for "otago AND auckland", no matching documents may include "otaio".
+            // Hopefully that made some sense...
+            if (count.intValue() > 0) {
+                System.out.println("  <Term value=\"" + term + "\" field=\"" + field + "\" freq=\"" + count.intValue() + "\" />");
+            }
             count = null;
             term = null;
             }
+
             // Cleanup
             terms_iter = null;