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
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
|
// Copyright Vespa.ai. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
#include <vespa/vespalib/testkit/test_kit.h>
#include <vespa/searchlib/features/setup.h>
#include <vespa/searchlib/fef/test/indexenvironment.h>
#include <vespa/searchlib/fef/test/indexenvironmentbuilder.h>
#include <vespa/searchlib/fef/test/queryenvironment.h>
#include <vespa/searchlib/features/text_similarity_feature.h>
#include <vespa/searchlib/fef/test/ftlib.h>
#include <initializer_list>
#include <vespa/searchlib/fef/test/dummy_dependency_handler.h>
#include <vespa/vespalib/util/stringfmt.h>
using namespace search::fef;
using namespace search::fef::test;
using namespace search::features;
using CollectionType = FieldInfo::CollectionType;
std::vector<vespalib::string> featureNamesFoo() {
std::vector<vespalib::string> f;
f.push_back("textSimilarity(foo).score");
f.push_back("textSimilarity(foo).proximity");
f.push_back("textSimilarity(foo).order");
f.push_back("textSimilarity(foo).queryCoverage");
f.push_back("textSimilarity(foo).fieldCoverage");
return f;
}
const size_t SCORE = 0;
const size_t PROXIMITY = 1;
const size_t ORDER = 2;
const size_t QUERY = 3;
const size_t FIELD = 4;
FtIndex indexFoo() {
FtIndex idx;
idx.field("foo");
return idx;
}
struct BlueprintFactoryFixture {
BlueprintFactory factory;
BlueprintFactoryFixture() : factory()
{
setup_search_features(factory);
}
};
struct IndexFixture {
IndexEnvironment indexEnv;
IndexFixture() : indexEnv()
{
IndexEnvironmentBuilder builder(indexEnv);
builder.addField(FieldType::INDEX, CollectionType::SINGLE, "foo");
builder.addField(FieldType::INDEX, CollectionType::WEIGHTEDSET, "bar");
builder.addField(FieldType::ATTRIBUTE, CollectionType::SINGLE, "baz");
}
};
struct FeatureDumpFixture : public IDumpFeatureVisitor {
std::vector<vespalib::string> expect;
size_t dumped;
virtual void visitDumpFeature(const vespalib::string &name) override {
EXPECT_LESS(dumped, expect.size());
EXPECT_EQUAL(expect[dumped++], name);
}
FeatureDumpFixture() : IDumpFeatureVisitor(), expect(featureNamesFoo()), dumped(0) {}
};
struct RankFixture : BlueprintFactoryFixture {
RankFixture() : BlueprintFactoryFixture() {}
double get_feature(const vespalib::string &query, const FtIndex &index, size_t select,
bool useStaleMatchData = false)
{
std::vector<vespalib::string> names = featureNamesFoo();
ASSERT_TRUE(names.size() == 5u);
FtFeatureTest ft(factory, names);
ft.getIndexEnv().getBuilder().addField(FieldType::INDEX, CollectionType::SINGLE, "foo");
FtTestApp::FT_SETUP(ft, FtUtil::toQuery(query), index, 1);
RankResult actual;
EXPECT_TRUE(ft.executeOnly(actual, useStaleMatchData ? 2 : 1));
return actual.getScore(names[select]);
}
};
double prox(uint32_t dist) {
return (dist > 8) ? 0 : (1.0 - (((dist-1)/8.0) * ((dist-1)/8.0)));
}
double comb(std::initializer_list<double> values) {
double sum = 0.0;
for (double value: values) {
sum += value;
}
return (sum/values.size());
}
double mix(double proximity, double order, double query, double field) {
return (0.35 * proximity) + (0.15 * order) + (0.30 * query) + (0.20 * field);
}
TEST_F("require that blueprint can be created from factory", BlueprintFactoryFixture) {
Blueprint::SP bp = f.factory.createBlueprint("textSimilarity");
EXPECT_TRUE(bp.get() != 0);
EXPECT_TRUE(dynamic_cast<TextSimilarityBlueprint*>(bp.get()) != 0);
}
TEST_FFF("require that appropriate features are dumped", TextSimilarityBlueprint, IndexFixture, FeatureDumpFixture) {
f1.visitDumpFeatures(f2.indexEnv, f3);
EXPECT_EQUAL(f3.expect.size(), f3.dumped);
}
TEST_FF("require that setup can be done on single value index field", TextSimilarityBlueprint, IndexFixture) {
DummyDependencyHandler deps(f1);
f1.setName(vespalib::make_string("%s(foo)", f1.getBaseName().c_str()));
EXPECT_TRUE(((Blueprint&)f1).setup(f2.indexEnv, std::vector<vespalib::string>(1, "foo")));
}
TEST_FF("require that setup can not be done on weighted set index field", TextSimilarityBlueprint, IndexFixture) {
DummyDependencyHandler deps(f1);
f1.setName(vespalib::make_string("%s(bar)", f1.getBaseName().c_str()));
EXPECT_TRUE(!((Blueprint&)f1).setup(f2.indexEnv, std::vector<vespalib::string>(1, "bar")));
}
TEST_FF("require that setup can not be done on single value attribute field", TextSimilarityBlueprint, IndexFixture) {
DummyDependencyHandler deps(f1);
f1.setName(vespalib::make_string("%s(baz)", f1.getBaseName().c_str()));
EXPECT_TRUE(!((Blueprint&)f1).setup(f2.indexEnv, std::vector<vespalib::string>(1, "baz")));
}
TEST_F("require that no match gives zero outputs", RankFixture) {
EXPECT_EQUAL(0.0, f1.get_feature("x", indexFoo().element("y"), SCORE));
EXPECT_EQUAL(0.0, f1.get_feature("x", indexFoo().element("y"), PROXIMITY));
EXPECT_EQUAL(0.0, f1.get_feature("x", indexFoo().element("y"), ORDER));
EXPECT_EQUAL(0.0, f1.get_feature("x", indexFoo().element("y"), QUERY));
EXPECT_EQUAL(0.0, f1.get_feature("x", indexFoo().element("y"), FIELD));
}
TEST_F("require that minal perfect match gives max outputs", RankFixture) {
EXPECT_EQUAL(1.0, f1.get_feature("x", indexFoo().element("x"), SCORE));
EXPECT_EQUAL(1.0, f1.get_feature("x", indexFoo().element("x"), PROXIMITY));
EXPECT_EQUAL(1.0, f1.get_feature("x", indexFoo().element("x"), ORDER));
EXPECT_EQUAL(1.0, f1.get_feature("x", indexFoo().element("x"), QUERY));
EXPECT_EQUAL(1.0, f1.get_feature("x", indexFoo().element("x"), FIELD));
}
TEST_F("require that larger perfect match gives max outputs", RankFixture) {
EXPECT_EQUAL(1.0, f1.get_feature("a b c d e f g", indexFoo().element("a b c d e f g"), SCORE));
EXPECT_EQUAL(1.0, f1.get_feature("a b c d e f g", indexFoo().element("a b c d e f g"), PROXIMITY));
EXPECT_EQUAL(1.0, f1.get_feature("a b c d e f g", indexFoo().element("a b c d e f g"), ORDER));
EXPECT_EQUAL(1.0, f1.get_feature("a b c d e f g", indexFoo().element("a b c d e f g"), QUERY));
EXPECT_EQUAL(1.0, f1.get_feature("a b c d e f g", indexFoo().element("a b c d e f g"), FIELD));
}
TEST_F("require that extra query terms reduces order but not proximity", RankFixture) {
EXPECT_EQUAL(1.0, f1.get_feature("x y", indexFoo().element("x"), PROXIMITY));
EXPECT_EQUAL(1.0, f1.get_feature("x y y", indexFoo().element("x"), PROXIMITY));
EXPECT_EQUAL(1.0, f1.get_feature("x y y y", indexFoo().element("x"), PROXIMITY));
EXPECT_EQUAL(0.0, f1.get_feature("x y", indexFoo().element("x"), ORDER));
EXPECT_EQUAL(0.0, f1.get_feature("x y y", indexFoo().element("x"), ORDER));
EXPECT_EQUAL(0.0, f1.get_feature("x y y y", indexFoo().element("x"), ORDER));
}
TEST_F("require that extra field terms reduces proximity but not order", RankFixture) {
EXPECT_EQUAL(prox(2), f1.get_feature("x", indexFoo().element("x y"), PROXIMITY));
EXPECT_EQUAL(prox(3), f1.get_feature("x", indexFoo().element("x y y"), PROXIMITY));
EXPECT_EQUAL(prox(4), f1.get_feature("x", indexFoo().element("x y y y"), PROXIMITY));
EXPECT_EQUAL(1.0, f1.get_feature("x", indexFoo().element("x y"), ORDER));
EXPECT_EQUAL(1.0, f1.get_feature("x", indexFoo().element("x y y"), ORDER));
EXPECT_EQUAL(1.0, f1.get_feature("x", indexFoo().element("x y y y"), ORDER));
}
TEST_F("require that proximity acts as expected", RankFixture) {
EXPECT_EQUAL(1.0, f1.get_feature("a b c d e", indexFoo().element("a b c d e"), PROXIMITY));
EXPECT_EQUAL(comb({prox(2), prox(1), prox(1), prox(1)}), f1.get_feature("a b c d e", indexFoo().element("a x b c d e"), PROXIMITY));
EXPECT_EQUAL(comb({prox(3), prox(1), prox(1), prox(1)}), f1.get_feature("a b c d e", indexFoo().element("a x x b c d e"), PROXIMITY));
EXPECT_EQUAL(comb({prox(4), prox(1), prox(1), prox(1)}), f1.get_feature("a b c d e", indexFoo().element("a x x x b c d e"), PROXIMITY));
EXPECT_EQUAL(comb({prox(2), prox(2), prox(2), prox(2)}), f1.get_feature("a b c d e", indexFoo().element("a x b x c x d x e"), PROXIMITY));
EXPECT_EQUAL(comb({prox(2), prox(2), prox(1), prox(3)}), f1.get_feature("a b c d e", indexFoo().element("a x b x c d x x e"), PROXIMITY));
}
TEST_F("require that field order does not affect proximity score", RankFixture) {
EXPECT_EQUAL(1.0, f1.get_feature("a b c d e", indexFoo().element("d c a b e"), PROXIMITY));
EXPECT_EQUAL(comb({prox(2), prox(1), prox(1), prox(1)}), f1.get_feature("a b c d e", indexFoo().element("d x c a b e"), PROXIMITY));
EXPECT_EQUAL(comb({prox(3), prox(1), prox(1), prox(1)}), f1.get_feature("a b c d e", indexFoo().element("d x x c a b e"), PROXIMITY));
EXPECT_EQUAL(comb({prox(4), prox(1), prox(1), prox(1)}), f1.get_feature("a b c d e", indexFoo().element("d x x x c a b e"), PROXIMITY));
EXPECT_EQUAL(comb({prox(2), prox(2), prox(2), prox(2)}), f1.get_feature("a b c d e", indexFoo().element("d x c x a x b x e"), PROXIMITY));
EXPECT_EQUAL(comb({prox(2), prox(2), prox(1), prox(3)}), f1.get_feature("a b c d e", indexFoo().element("d x c x a b x x e"), PROXIMITY));
}
TEST_F("require that order score acts as expected", RankFixture) {
EXPECT_EQUAL(1.0, f1.get_feature("a b c d e", indexFoo().element("a b c d e"), ORDER));
EXPECT_EQUAL(comb({1.0, 1.0, 1.0, 0.0}), f1.get_feature("a b c d e", indexFoo().element("a b c e d"), ORDER));
EXPECT_EQUAL(comb({0.0, 1.0, 1.0, 0.0}), f1.get_feature("a b c d e", indexFoo().element("b a c e d"), ORDER));
EXPECT_EQUAL(comb({0.0, 1.0, 0.0, 0.0}), f1.get_feature("a b c d e", indexFoo().element("b a e d c"), ORDER));
EXPECT_EQUAL(comb({0.0, 0.0, 0.0, 0.0}), f1.get_feature("a b c d e", indexFoo().element("e d c b a"), ORDER));
}
TEST_F("require that proximity does not affect order score", RankFixture) {
EXPECT_EQUAL(1.0, f1.get_feature("a b c d e", indexFoo().element("a b c d e"), ORDER));
EXPECT_EQUAL(comb({1.0, 1.0, 1.0, 0.0}), f1.get_feature("a b c d e", indexFoo().element("a x b x c x e x d"), ORDER));
EXPECT_EQUAL(comb({0.0, 1.0, 1.0, 0.0}), f1.get_feature("a b c d e", indexFoo().element("b x a x c x e x d"), ORDER));
EXPECT_EQUAL(comb({0.0, 1.0, 0.0, 0.0}), f1.get_feature("a b c d e", indexFoo().element("b x a x e x d x c"), ORDER));
EXPECT_EQUAL(comb({0.0, 0.0, 0.0, 0.0}), f1.get_feature("a b c d e", indexFoo().element("e x d x c x b x a"), ORDER));
}
TEST_F("require that query coverage acts as expected", RankFixture) {
EXPECT_EQUAL(5.0/5.0, f1.get_feature("a b c d e", indexFoo().element("a b c d e"), QUERY));
EXPECT_EQUAL(4.0/5.0, f1.get_feature("a b c d e", indexFoo().element("a b c d"), QUERY));
EXPECT_EQUAL(3.0/5.0, f1.get_feature("a b c d e", indexFoo().element("a b c"), QUERY));
EXPECT_EQUAL(2.0/5.0, f1.get_feature("a b c d e", indexFoo().element("a b"), QUERY));
EXPECT_EQUAL(4.0/7.0, f1.get_feature("a!200 b!200 c d e", indexFoo().element("a b"), QUERY));
EXPECT_EQUAL(2.0/7.0, f1.get_feature("a b c!500", indexFoo().element("a b"), QUERY));
EXPECT_EQUAL(5.0/7.0, f1.get_feature("a b c!500", indexFoo().element("c"), QUERY));
}
TEST_F("require that field coverage acts as expected", RankFixture) {
EXPECT_EQUAL(5.0/5.0, f1.get_feature("a b c d e", indexFoo().element("a b c d e"), FIELD));
EXPECT_EQUAL(4.0/5.0, f1.get_feature("a b c d e", indexFoo().element("a x c d e"), FIELD));
EXPECT_EQUAL(3.0/5.0, f1.get_feature("a b c d e", indexFoo().element("a b x x e"), FIELD));
EXPECT_EQUAL(2.0/5.0, f1.get_feature("a b c d e", indexFoo().element("x x x d e"), FIELD));
}
TEST_F("require that first unique match is used per query term", RankFixture) {
EXPECT_EQUAL(prox(3), f1.get_feature("a b", indexFoo().element("a a a b"), PROXIMITY));
EXPECT_EQUAL(1.0, f1.get_feature("a b", indexFoo().element("a a a b"), ORDER));
EXPECT_EQUAL(1.0, f1.get_feature("a b", indexFoo().element("a a a b"), QUERY));
EXPECT_EQUAL(2.0/4.0, f1.get_feature("a b", indexFoo().element("a a a b"), FIELD));
EXPECT_EQUAL(comb({prox(1), prox(2)}), f1.get_feature("a b a", indexFoo().element("a a a b"), PROXIMITY));
EXPECT_EQUAL(0.5, f1.get_feature("a b a", indexFoo().element("a a a b"), ORDER));
EXPECT_EQUAL(1.0, f1.get_feature("a b a", indexFoo().element("a a a b"), QUERY));
EXPECT_EQUAL(3.0/4.0, f1.get_feature("a b a", indexFoo().element("a a a b"), FIELD));
}
TEST_F("require that overall score combines individual signals appropriately", RankFixture) {
EXPECT_EQUAL(comb({prox(1), prox(3), prox(2)}), f1.get_feature("a b c d e", indexFoo().element("a c x x b x d"), PROXIMITY));
EXPECT_EQUAL(comb({1.0, 0.0, 1.0}), f1.get_feature("a b c d e", indexFoo().element("a c x x b x d"), ORDER));
EXPECT_EQUAL(4.0/5.0, f1.get_feature("a b c d e", indexFoo().element("a c x x b x d"), QUERY));
EXPECT_EQUAL(4.0/7.0, f1.get_feature("a b c d e", indexFoo().element("a c x x b x d"), FIELD));
EXPECT_EQUAL(mix(comb({prox(1), prox(3), prox(2)}), comb({1.0, 0.0, 1.0}), 4.0/5.0, 4.0/7.0),
f1.get_feature("a b c d e", indexFoo().element("a c x x b x d"), SCORE));
}
TEST_MAIN() { TEST_RUN_ALL(); }
|
