Author: tommaso
Date: Tue Mar 12 13:20:14 2019
New Revision: 1855318
URL: http://svn.apache.org/viewvc?rev=1855318&view=rev
Log:
OAK-8119 - use exact distance scoring in reranking (branch 1.10)
Modified:
jackrabbit/oak/branches/1.10/ (props changed)
jackrabbit/oak/branches/1.10/oak-lucene/src/main/java/org/apache/jackrabbit/oak/plugins/index/lucene/util/fv/SimSearchUtils.java
Propchange: jackrabbit/oak/branches/1.10/
------------------------------------------------------------------------------
--- svn:mergeinfo (original)
+++ svn:mergeinfo Tue Mar 12 13:20:14 2019
@@ -1,3 +1,3 @@
/jackrabbit/oak/branches/1.0:1665962
-/jackrabbit/oak/trunk:1850874,1850882,1851236,1851253,1851451,1851533-1851535,1851619,1852052,1852084,1852120,1852451,1852492-1852493,1852528,1852582,1852584,1852601,1852920,1853141,1853229,1853393,1853429,1853433,1853441,1853866,1853868,1853870,1853893,1853969,1853997,1854034,1854044,1854058,1854113,1854373,1854377,1854380,1854385,1854401,1854403,1854455,1854461-1854462,1854466,1854468,1854515,1854533,1854701,1854827,1854848,1854859,1854930,1854990,1855221
+/jackrabbit/oak/trunk:1850874,1850882,1851236,1851253,1851451,1851533-1851535,1851619,1852052,1852084,1852120,1852451,1852492-1852493,1852528,1852582,1852584,1852601,1852920,1853141,1853229,1853393,1853429,1853433,1853441,1853866,1853868,1853870,1853893,1853969,1853997,1854034,1854044,1854058,1854113,1854373,1854377,1854380,1854385,1854401,1854403,1854455,1854461-1854462,1854466,1854468,1854515,1854533,1854701,1854827,1854848,1854859,1854930,1854990-1854991,1855221
/jackrabbit/trunk:1345480
Modified:
jackrabbit/oak/branches/1.10/oak-lucene/src/main/java/org/apache/jackrabbit/oak/plugins/index/lucene/util/fv/SimSearchUtils.java
URL:
http://svn.apache.org/viewvc/jackrabbit/oak/branches/1.10/oak-lucene/src/main/java/org/apache/jackrabbit/oak/plugins/index/lucene/util/fv/SimSearchUtils.java?rev=1855318&r1=1855317&r2=1855318&view=diff
==============================================================================
---
jackrabbit/oak/branches/1.10/oak-lucene/src/main/java/org/apache/jackrabbit/oak/plugins/index/lucene/util/fv/SimSearchUtils.java
(original)
+++
jackrabbit/oak/branches/1.10/oak-lucene/src/main/java/org/apache/jackrabbit/oak/plugins/index/lucene/util/fv/SimSearchUtils.java
Tue Mar 12 13:20:14 2019
@@ -255,16 +255,19 @@ public class SimSearchUtils {
if (binaryValue != null) {
double[] inputVector = toDoubleArray(binaryValue.bytes);
for (int j = 0; j < docs.scoreDocs.length; j++) {
- double[] currentVector =
toDoubleArray(indexSearcher.doc(docs.scoreDocs[j].doc)
- .getBinaryValue(fieldName).bytes);
- double distance = dist(inputVector, currentVector) +
1e-10; // constant term to avoid division by zero
- if (Double.isNaN(distance) || Double.isInfinite(distance))
{
- toDiscard.add(docs.scoreDocs[j].doc);
- } else {
- distSum += distance;
- counter++;
- distances.put(docs.scoreDocs[j].doc, distance);
- docs.scoreDocs[j].score += (float) (1d / distance); //
additive similarity boosting
+ BytesRef featureVectorBinary =
indexSearcher.doc(docs.scoreDocs[j].doc)
+ .getBinaryValue(fieldName);
+ if (featureVectorBinary != null) {
+ double[] currentVector =
toDoubleArray(featureVectorBinary.bytes);
+ double distance = dist(inputVector, currentVector) +
1e-10; // constant term to avoid division by zero
+ if (Double.isNaN(distance) ||
Double.isInfinite(distance)) {
+ toDiscard.add(docs.scoreDocs[j].doc);
+ } else {
+ distSum += distance;
+ counter++;
+ distances.put(docs.scoreDocs[j].doc, distance);
+ docs.scoreDocs[j].score = (float) (1d / distance);
+ }
}
}
}