Apache
/
lucene
mirror of https://github.com/apache/lucene.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

LUCENE-4133: FVH: A weighted approach for ordered fragments, part of LUCENE-3440 

git-svn-id: https://svn.apache.org/repos/asf/lucene/dev/trunk@1349361 13f79535-47bb-0310-9956-ffa450edef68
This commit is contained in:
Koji Sekiguchi 2012-06-12 13:59:37 +00:00
parent 4b212a7c4f
commit 2210749fbd
7 changed files with 218 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
lucene/CHANGES.txt
View File

@ -904,6 +904,9 @@ New features
cause a ParseException (depending on whether strict parsing is enabled).
(Luca Cavanna via Chris Male)
* LUCENE-3440: Add ordered fragments feature with IDF-weighted terms for FVH.
(Sebastian Lutze via Koji Sekiguchi)
Optimizations
* LUCENE-2588: Don't store unnecessary suffixes when writing the terms

4
lucene/highlighter/src/java/org/apache/lucene/search/vectorhighlight/FieldPhraseList.java
View File

@ -150,7 +150,7 @@ public class FieldPhraseList {
}
/**
* @return the termInfos
* @return the termInfos
*/
public List<TermInfo> getTermsInfos() {
return termsInfos;
@ -164,7 +164,7 @@ public class FieldPhraseList {
this.boost = boost;
this.seqnum = seqnum;
// now we keep TermInfos for further operations
// We keep TermInfos for further operations
termsInfos = new ArrayList<TermInfo>( terms );
termsOffsets = new ArrayList<Toffs>( terms.size() );

7
lucene/highlighter/src/java/org/apache/lucene/search/vectorhighlight/SimpleFieldFragList.java
View File

@ -42,12 +42,13 @@ public class SimpleFieldFragList extends FieldFragList {
*/
@Override
public void add( int startOffset, int endOffset, List<WeightedPhraseInfo> phraseInfoList ) {
float score = 0;
float totalBoost = 0;
List<SubInfo> subInfos = new ArrayList<SubInfo>();
for( WeightedPhraseInfo phraseInfo : phraseInfoList ){
subInfos.add( new SubInfo( phraseInfo.getText(), phraseInfo.getTermsOffsets(), phraseInfo.getSeqnum() ) );
score += phraseInfo.getBoost();
totalBoost += phraseInfo.getBoost();
}
getFragInfos().add( new WeightedFragInfo( startOffset, endOffset, subInfos, score ) );
getFragInfos().add( new WeightedFragInfo( startOffset, endOffset, subInfos, totalBoost ) );
}
}

76
lucene/highlighter/src/java/org/apache/lucene/search/vectorhighlight/WeightedFieldFragList.java Normal file
View File

@ -0,0 +1,76 @@
package org.apache.lucene.search.vectorhighlight;
/**
* Licensed to the Apache Software Foundation (ASF) under one or more
* contributor license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright ownership.
* The ASF licenses this file to You under the Apache License, Version 2.0
* (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
import java.util.ArrayList;
import java.util.HashSet;
import java.util.List;
import org.apache.lucene.search.vectorhighlight.FieldFragList.WeightedFragInfo.SubInfo;
import org.apache.lucene.search.vectorhighlight.FieldPhraseList.WeightedPhraseInfo;
import org.apache.lucene.search.vectorhighlight.FieldTermStack.TermInfo;
/**
* A weighted implementation of {@link FieldFragList}.
*/
public class WeightedFieldFragList extends FieldFragList {
/**
* a constructor.
*
* @param fragCharSize the length (number of chars) of a fragment
*/
public WeightedFieldFragList( int fragCharSize ) {
super( fragCharSize );
}
/* (non-Javadoc)
* @see org.apache.lucene.search.vectorhighlight.FieldFragList#add( int startOffset, int endOffset, List<WeightedPhraseInfo> phraseInfoList )
*/
@Override
public void add( int startOffset, int endOffset, List<WeightedPhraseInfo> phraseInfoList ) {
float totalBoost = 0;
List<SubInfo> subInfos = new ArrayList<SubInfo>();
HashSet<String> distinctTerms = new HashSet<String>();
int length = 0;
for( WeightedPhraseInfo phraseInfo : phraseInfoList ){
subInfos.add( new SubInfo( phraseInfo.getText(), phraseInfo.getTermsOffsets(), phraseInfo.getSeqnum() ) );
for ( TermInfo ti : phraseInfo.getTermsInfos()) {
if ( distinctTerms.add( ti.getText() ) )
totalBoost += ti.getWeight() * phraseInfo.getBoost();
length++;
}
}
// We want that terms per fragment (length) is included into the weight. Otherwise a one-word-query
// would cause an equal weight for all fragments regardless of how much words they contain.
// To avoid that fragments containing a high number of words possibly "outrank" more relevant fragments
// we "bend" the length with a standard-normalization a little bit.
totalBoost *= length * ( 1 / Math.sqrt( length ) );
getFragInfos().add( new WeightedFragInfo( startOffset, endOffset, subInfos, totalBoost ) );
}
}

41
lucene/highlighter/src/java/org/apache/lucene/search/vectorhighlight/WeightedFragListBuilder.java Normal file
View File

@ -0,0 +1,41 @@
package org.apache.lucene.search.vectorhighlight;
/**
* Licensed to the Apache Software Foundation (ASF) under one or more
* contributor license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright ownership.
* The ASF licenses this file to You under the Apache License, Version 2.0
* (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
/**
* A weighted implementation of {@link FragListBuilder}.
*/
public class WeightedFragListBuilder extends BaseFragListBuilder {
public WeightedFragListBuilder() {
super();
}
public WeightedFragListBuilder(int margin) {
super(margin);
}
/* (non-Javadoc)
* @see org.apache.lucene.search.vectorhighlight.FragListBuilder#createFieldFragList(FieldPhraseList fieldPhraseList, int fragCharSize)
*/
@Override
public FieldFragList createFieldFragList( FieldPhraseList fieldPhraseList, int fragCharSize ){
return createFieldFragList( fieldPhraseList, new WeightedFieldFragList( fragCharSize ), fragCharSize );
}
}

62
lucene/highlighter/src/java/org/apache/lucene/search/vectorhighlight/package.html
View File

@ -27,9 +27,9 @@ This is an another highlighter implementation.
<li>support multi-term (includes wildcard, range, regexp, etc) queries</li>
<li>need Java 1.5</li>
<li>highlight fields need to be stored with Positions and Offsets</li>
<li>take into account query boost to score fragments</li>
<li>take into account query boost and/or IDF-weight to score fragments</li>
<li>support colored highlight tags</li>
<li>pluggable FragListBuilder</li>
<li>pluggable FragListBuilder / FieldFragList</li>
<li>pluggable FragmentsBuilder</li>
</ul>
@ -122,9 +122,8 @@ by reference to <code>QueryPhraseMap</code> and <code>FieldTermStack</code>.</p>
+----------------+-----------------+---+
</pre>
<p>The type of each entry is <code>WeightedPhraseInfo</code> that consists of
an array of terms offsets and weight. The weight (Fast Vector Highlighter uses query boost to
calculate the weight) will be taken into account when Fast Vector Highlighter creates
{@link org.apache.lucene.search.vectorhighlight.FieldFragList} in the next step.</p>
an array of terms offsets and weight.
</p>
<h3>Step 4.</h3>
<p>In Step 4, Fast Vector Highlighter creates <code>FieldFragList</code> by reference to
<code>FieldPhraseList</code>. In this sample case, the following
@ -137,6 +136,59 @@ calculate the weight) will be taken into account when Fast Vector Highlighter cr
|totalBoost=3 |
+---------------------------------+
</pre>
<p>
The calculation for each <code>FieldFragList.WeightedFragInfo.totalBoost</code> (weight)
depends on the implementation of <code>FieldFragList.add( ... )</code>:
<pre class="prettyprint">
public void add( int startOffset, int endOffset, List&lt;WeightedPhraseInfo&gt; phraseInfoList ) {
float totalBoost = 0;
List&lt;SubInfo&gt; subInfos = new ArrayList&lt;SubInfo&gt;();
for( WeightedPhraseInfo phraseInfo : phraseInfoList ){
subInfos.add( new SubInfo( phraseInfo.getText(), phraseInfo.getTermsOffsets(), phraseInfo.getSeqnum() ) );
totalBoost += phraseInfo.getBoost();
}
getFragInfos().add( new WeightedFragInfo( startOffset, endOffset, subInfos, totalBoost ) );
}
</pre>
The used implementation of <code>FieldFragList</code> is noted in <code>BaseFragListBuilder.createFieldFragList( ... )</code>:
<pre class="prettyprint">
public FieldFragList createFieldFragList( FieldPhraseList fieldPhraseList, int fragCharSize ){
return createFieldFragList( fieldPhraseList, new SimpleFieldFragList( fragCharSize ), fragCharSize );
}
</pre>
<p>
Currently there are basically to approaches available:
</p>
<ul>
<li><code>SimpleFragListBuilder using SimpleFieldFragList</code>: <i>sum-of-boosts</i>-approach. The totalBoost is calculated by summarizing the query-boosts per term. Per default a term is boosted by 1.0</li>
<li><code>WeightedFragListBuilder using WeightedFieldFragList</code>: <i>sum-of-distinct-weights</i>-approach. The totalBoost is calculated by summarizing the IDF-weights of distinct terms.</li>
</ul>
<p>Comparison of the two approaches:</p>
<table border="1">
<caption>
query = das alte testament (The Old Testament)
</caption>
<tr><th>Terms in fragment</th><th>sum-of-distinct-weights</th><th>sum-of-boosts</th></tr>
<tr><td>das alte testament</td><td>5.339621</td><td>3.0</td></tr>
<tr><td>das alte testament</td><td>5.339621</td><td>3.0</td></tr>
<tr><td>das testament alte</td><td>5.339621</td><td>3.0</td></tr>
<tr><td>das alte testament</td><td>5.339621</td><td>3.0</td></tr>
<tr><td>das testament</td><td>2.9455688</td><td>2.0</td></tr>
<tr><td>das alte</td><td>2.4759595</td><td>2.0</td></tr>
<tr><td>das das das das</td><td>1.5015357</td><td>4.0</td></tr>
<tr><td>das das das</td><td>1.3003681</td><td>3.0</td></tr>
<tr><td>das das</td><td>1.061746</td><td>2.0</td></tr>
<tr><td>alte</td><td>1.0</td><td>1.0</td></tr>
<tr><td>alte</td><td>1.0</td><td>1.0</td></tr>
<tr><td>das</td><td>0.7507678</td><td>1.0</td></tr>
<tr><td>das</td><td>0.7507678</td><td>1.0</td></tr>
<tr><td>das</td><td>0.7507678</td><td>1.0</td></tr>
<tr><td>das</td><td>0.7507678</td><td>1.0</td></tr>
<tr><td>das</td><td>0.7507678</td><td>1.0</td></tr>
</table>
<h3>Step 5.</h3>
<p>In Step 5, by using <code>FieldFragList</code> and the field stored data,
Fast Vector Highlighter creates highlighted snippets!</p>

35
lucene/highlighter/src/test/org/apache/lucene/search/vectorhighlight/WeightedFragListBuilderTest.java Normal file
View File

@ -0,0 +1,35 @@
package org.apache.lucene.search.vectorhighlight;
/**
* Licensed to the Apache Software Foundation (ASF) under one or more
* contributor license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright ownership.
* The ASF licenses this file to You under the Apache License, Version 2.0
* (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
public class WeightedFragListBuilderTest extends AbstractTestCase {
public void test2WeightedFragList() throws Exception {
makeIndexLongMV();
FieldQuery fq = new FieldQuery( pqF( "the", "both" ), true, true );
FieldTermStack stack = new FieldTermStack( reader, 0, F, fq );
FieldPhraseList fpl = new FieldPhraseList( stack, fq );
WeightedFragListBuilder wflb = new WeightedFragListBuilder();
FieldFragList ffl = wflb.createFieldFragList( fpl, 100 );
assertEquals( 1, ffl.getFragInfos().size() );
assertEquals( "subInfos=(theboth((195,203)))/0.86791086(189,289)", ffl.getFragInfos().get( 0 ).toString() );
}
}