Package org.apache.solr.ltr

Class LTRRescorer

java.lang.Object

org.apache.lucene.search.Rescorer

org.apache.solr.ltr.LTRRescorer

Direct Known Subclasses:

LTRInterleavingRescorer

public class LTRRescorer
extends org.apache.lucene.search.Rescorer

Implements the rescoring logic. The top documents returned by solr with their original scores, will be processed by a LTRScoringQuery that will assign a new score to each document. The top documents will be resorted based on the new score.

Field Summary

Fields

Modifier and Type	Field	Description
protected static Comparator<org.apache.lucene.search.ScoreDoc>	scoreComparator

Constructor Summary

Constructors

Constructor	Description
LTRRescorer()
LTRRescorer(LTRScoringQuery scoringQuery)

Method Summary

Modifier and Type	Method	Description
org.apache.lucene.search.Explanation	explain(org.apache.lucene.search.IndexSearcher searcher, org.apache.lucene.search.Explanation firstPassExplanation, int docID)
static LTRScoringQuery.FeatureInfo[]	extractFeaturesInfo(LTRScoringQuery.ModelWeight modelWeight, int docid, Float originalDocScore, List<org.apache.lucene.index.LeafReaderContext> leafContexts)
protected static org.apache.lucene.search.Explanation	getExplanation(org.apache.lucene.search.IndexSearcher searcher, int docID, LTRScoringQuery rerankingQuery)
protected static org.apache.lucene.search.ScoreDoc[]	getFirstPassDocsRanked(org.apache.lucene.search.TopDocs firstPassTopDocs)
protected static void	heapAdjust(org.apache.lucene.search.ScoreDoc[] hits, int size, int root)
protected static void	heapify(org.apache.lucene.search.ScoreDoc[] hits, int size)
protected static void	logSingleHit(org.apache.lucene.search.IndexSearcher indexSearcher, LTRScoringQuery.ModelWeight modelWeight, int docid, LTRScoringQuery scoringQuery)	Call this method if the scoreSingleHit(int, int, int, ScoreDoc, int, org.apache.solr.ltr.LTRScoringQuery.ModelWeight.ModelScorer, ScoreDoc[]) method indicated that the document's feature info should be logged.
org.apache.lucene.search.TopDocs	rescore(org.apache.lucene.search.IndexSearcher searcher, org.apache.lucene.search.TopDocs firstPassTopDocs, int topN)	rescores the documents:
void	scoreFeatures(org.apache.lucene.search.IndexSearcher indexSearcher, int topN, LTRScoringQuery.ModelWeight modelWeight, org.apache.lucene.search.ScoreDoc[] hits, List<org.apache.lucene.index.LeafReaderContext> leaves, org.apache.lucene.search.ScoreDoc[] reranked)
protected static boolean	scoreSingleHit(int topN, int docBase, int hitUpto, org.apache.lucene.search.ScoreDoc hit, int docID, LTRScoringQuery.ModelWeight.ModelScorer scorer, org.apache.lucene.search.ScoreDoc[] reranked)	Scores a single document and returns true if the document's feature info should be logged via the logSingleHit(IndexSearcher, org.apache.solr.ltr.LTRScoringQuery.ModelWeight, int, LTRScoringQuery) method.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

scoreComparator

protected static final Comparator<org.apache.lucene.search.ScoreDoc> scoreComparator

Constructor Detail

LTRRescorer

public LTRRescorer()

LTRRescorer

public LTRRescorer(LTRScoringQuery scoringQuery)

Method Detail

heapAdjust

protected static void heapAdjust(org.apache.lucene.search.ScoreDoc[] hits,
                                 int size,
                                 int root)

heapify

protected static void heapify(org.apache.lucene.search.ScoreDoc[] hits,
                              int size)

rescore

public org.apache.lucene.search.TopDocs rescore(org.apache.lucene.search.IndexSearcher searcher,
                                                org.apache.lucene.search.TopDocs firstPassTopDocs,
                                                int topN)
                                         throws IOException

rescores the documents:

Specified by:

rescore in class org.apache.lucene.search.Rescorer

Parameters:

searcher - current IndexSearcher

firstPassTopDocs - documents to rerank;

topN - documents to return;

Throws:

IOException

getFirstPassDocsRanked

protected static org.apache.lucene.search.ScoreDoc[] getFirstPassDocsRanked(org.apache.lucene.search.TopDocs firstPassTopDocs)

scoreFeatures

public void scoreFeatures(org.apache.lucene.search.IndexSearcher indexSearcher,
                          int topN,
                          LTRScoringQuery.ModelWeight modelWeight,
                          org.apache.lucene.search.ScoreDoc[] hits,
                          List<org.apache.lucene.index.LeafReaderContext> leaves,
                          org.apache.lucene.search.ScoreDoc[] reranked)
                   throws IOException

Throws:

IOException

logSingleHit

protected static void logSingleHit(org.apache.lucene.search.IndexSearcher indexSearcher,
                                   LTRScoringQuery.ModelWeight modelWeight,
                                   int docid,
                                   LTRScoringQuery scoringQuery)

Call this method if the scoreSingleHit(int, int, int, ScoreDoc, int, org.apache.solr.ltr.LTRScoringQuery.ModelWeight.ModelScorer, ScoreDoc[]) method indicated that the document's feature info should be logged.

scoreSingleHit

protected static boolean scoreSingleHit(int topN,
                                        int docBase,
                                        int hitUpto,
                                        org.apache.lucene.search.ScoreDoc hit,
                                        int docID,
                                        LTRScoringQuery.ModelWeight.ModelScorer scorer,
                                        org.apache.lucene.search.ScoreDoc[] reranked)
                                 throws IOException

Scores a single document and returns true if the document's feature info should be logged via the logSingleHit(IndexSearcher, org.apache.solr.ltr.LTRScoringQuery.ModelWeight, int, LTRScoringQuery) method. Feature info logging is only necessary for the topN documents.

Throws:

IOException

explain

public org.apache.lucene.search.Explanation explain(org.apache.lucene.search.IndexSearcher searcher,
                                                    org.apache.lucene.search.Explanation firstPassExplanation,
                                                    int docID)
                                             throws IOException

Specified by:

explain in class org.apache.lucene.search.Rescorer

Throws:

IOException

getExplanation

protected static org.apache.lucene.search.Explanation getExplanation(org.apache.lucene.search.IndexSearcher searcher,
                                                                     int docID,
                                                                     LTRScoringQuery rerankingQuery)
                                                              throws IOException

Throws:

IOException

extractFeaturesInfo

public static LTRScoringQuery.FeatureInfo[] extractFeaturesInfo(LTRScoringQuery.ModelWeight modelWeight,
                                                                int docid,
                                                                Float originalDocScore,
                                                                List<org.apache.lucene.index.LeafReaderContext> leafContexts)
                                                         throws IOException

Throws:

IOException