/
QueryRescorer.java
207 lines (171 loc) · 8.09 KB
/
QueryRescorer.java
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
/*
* Copyright Elasticsearch B.V. and/or licensed to Elasticsearch B.V. under one
* or more contributor license agreements. Licensed under the Elastic License
* 2.0 and the Server Side Public License, v 1; you may not use this file except
* in compliance with, at your election, the Elastic License 2.0 or the Server
* Side Public License, v 1.
*/
package org.elasticsearch.search.rescore;
import org.apache.lucene.search.Explanation;
import org.apache.lucene.search.IndexSearcher;
import org.apache.lucene.search.ScoreDoc;
import org.apache.lucene.search.TopDocs;
import org.elasticsearch.index.query.ParsedQuery;
import java.io.IOException;
import java.util.Arrays;
import java.util.Collections;
import java.util.Comparator;
import java.util.List;
import java.util.Set;
import static java.util.stream.Collectors.toSet;
public final class QueryRescorer implements Rescorer {
public static final Rescorer INSTANCE = new QueryRescorer();
@Override
public TopDocs rescore(TopDocs topDocs, IndexSearcher searcher, RescoreContext rescoreContext) throws IOException {
assert rescoreContext != null;
if (topDocs == null || topDocs.scoreDocs.length == 0) {
return topDocs;
}
final QueryRescoreContext rescore = (QueryRescoreContext) rescoreContext;
org.apache.lucene.search.Rescorer rescorer = new org.apache.lucene.search.QueryRescorer(rescore.parsedQuery().query()) {
@Override
protected float combine(float firstPassScore, boolean secondPassMatches, float secondPassScore) {
if (secondPassMatches) {
return rescore.scoreMode.combine(
firstPassScore * rescore.queryWeight(),
secondPassScore * rescore.rescoreQueryWeight()
);
}
// TODO: shouldn't this be up to the ScoreMode? I.e., we should just invoke ScoreMode.combine, passing 0.0f for the
// secondary score?
return firstPassScore * rescore.queryWeight();
}
};
// First take top slice of incoming docs, to be rescored:
TopDocs topNFirstPass = topN(topDocs, rescoreContext.getWindowSize());
// Save doc IDs for which rescoring was applied to be used in score explanation
Set<Integer> topNDocIDs = Collections.unmodifiableSet(
Arrays.stream(topNFirstPass.scoreDocs).map(scoreDoc -> scoreDoc.doc).collect(toSet())
);
rescoreContext.setRescoredDocs(topNDocIDs);
// Rescore them:
TopDocs rescored = rescorer.rescore(searcher, topNFirstPass, rescoreContext.getWindowSize());
// Splice back to non-topN hits and resort all of them:
return combine(topDocs, rescored, (QueryRescoreContext) rescoreContext);
}
@Override
public Explanation explain(int topLevelDocId, IndexSearcher searcher, RescoreContext rescoreContext, Explanation sourceExplanation)
throws IOException {
if (sourceExplanation == null) {
// this should not happen but just in case
return Explanation.noMatch("nothing matched");
}
QueryRescoreContext rescore = (QueryRescoreContext) rescoreContext;
float primaryWeight = rescore.queryWeight();
Explanation prim;
if (sourceExplanation.isMatch()) {
prim = Explanation.match(
sourceExplanation.getValue().floatValue() * primaryWeight,
"product of:",
sourceExplanation,
Explanation.match(primaryWeight, "primaryWeight")
);
} else {
prim = Explanation.noMatch("First pass did not match", sourceExplanation);
}
if (rescoreContext.isRescored(topLevelDocId)) {
Explanation rescoreExplain = searcher.explain(rescore.parsedQuery().query(), topLevelDocId);
// NOTE: we don't use Lucene's Rescorer.explain because we want to insert our own description with which ScoreMode was used.
// Maybe we should add QueryRescorer.explainCombine to Lucene?
if (rescoreExplain != null && rescoreExplain.isMatch()) {
float secondaryWeight = rescore.rescoreQueryWeight();
Explanation sec = Explanation.match(
rescoreExplain.getValue().floatValue() * secondaryWeight,
"product of:",
rescoreExplain,
Explanation.match(secondaryWeight, "secondaryWeight")
);
QueryRescoreMode scoreMode = rescore.scoreMode();
return Explanation.match(
scoreMode.combine(prim.getValue().floatValue(), sec.getValue().floatValue()),
scoreMode + " of:",
prim,
sec
);
}
}
return prim;
}
private static final Comparator<ScoreDoc> SCORE_DOC_COMPARATOR = (o1, o2) -> {
int cmp = Float.compare(o2.score, o1.score);
return cmp == 0 ? Integer.compare(o1.doc, o2.doc) : cmp;
};
/** Returns a new {@link TopDocs} with the topN from the incoming one, or the same TopDocs if the number of hits is already <=
* topN. */
private static TopDocs topN(TopDocs in, int topN) {
if (in.scoreDocs.length < topN) {
return in;
}
ScoreDoc[] subset = new ScoreDoc[topN];
System.arraycopy(in.scoreDocs, 0, subset, 0, topN);
return new TopDocs(in.totalHits, subset);
}
/** Modifies incoming TopDocs (in) by replacing the top hits with resorted's hits, and then resorting all hits. */
private static TopDocs combine(TopDocs in, TopDocs resorted, QueryRescoreContext ctx) {
System.arraycopy(resorted.scoreDocs, 0, in.scoreDocs, 0, resorted.scoreDocs.length);
if (in.scoreDocs.length > resorted.scoreDocs.length) {
// These hits were not rescored (beyond the rescore window), so we treat them the same as a hit that did get rescored but did
// not match the 2nd pass query:
for (int i = resorted.scoreDocs.length; i < in.scoreDocs.length; i++) {
// TODO: shouldn't this be up to the ScoreMode? I.e., we should just invoke ScoreMode.combine, passing 0.0f for the
// secondary score?
in.scoreDocs[i].score *= ctx.queryWeight();
}
// TODO: this is wrong, i.e. we are comparing apples and oranges at this point. It would be better if we always rescored all
// incoming first pass hits, instead of allowing recoring of just the top subset:
Arrays.sort(in.scoreDocs, SCORE_DOC_COMPARATOR);
}
return in;
}
public static class QueryRescoreContext extends RescoreContext {
private ParsedQuery query;
private float queryWeight = 1.0f;
private float rescoreQueryWeight = 1.0f;
private QueryRescoreMode scoreMode;
public QueryRescoreContext(int windowSize) {
super(windowSize, QueryRescorer.INSTANCE);
this.scoreMode = QueryRescoreMode.Total;
}
public void setQuery(ParsedQuery query) {
this.query = query;
}
@Override
public List<ParsedQuery> getParsedQueries() {
return Collections.singletonList(query);
}
public ParsedQuery parsedQuery() {
return query;
}
public float queryWeight() {
return queryWeight;
}
public float rescoreQueryWeight() {
return rescoreQueryWeight;
}
public QueryRescoreMode scoreMode() {
return scoreMode;
}
public void setRescoreQueryWeight(float rescoreQueryWeight) {
this.rescoreQueryWeight = rescoreQueryWeight;
}
public void setQueryWeight(float queryWeight) {
this.queryWeight = queryWeight;
}
public void setScoreMode(QueryRescoreMode scoreMode) {
this.scoreMode = scoreMode;
}
public void setScoreMode(String scoreMode) {
setScoreMode(QueryRescoreMode.fromString(scoreMode));
}
}
}