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
|
// Copyright 2017 Yahoo Holdings. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
#include "nativeproximityfeature.h"
#include "valuefeature.h"
#include "utils.h"
#include <vespa/searchlib/fef/fieldinfo.h>
#include <vespa/searchlib/fef/indexproperties.h>
#include <vespa/searchlib/fef/itablemanager.h>
#include <vespa/searchlib/fef/properties.h>
#include <map>
using namespace search::fef;
namespace search {
namespace features {
feature_t
NativeProximityExecutor::calculateScoreForField(const FieldSetup & fs, uint32_t docId)
{
feature_t score = 0;
for (size_t i = 0; i < fs.pairs.size(); ++i) {
score += calculateScoreForPair(fs.pairs[i], fs.fieldId, docId);
}
score *= _params.vector[fs.fieldId].fieldWeight;
if (fs.divisor > 0) {
score /= fs.divisor;
}
return score;
}
feature_t
NativeProximityExecutor::calculateScoreForPair(const TermPair & pair, uint32_t fieldId, uint32_t docId)
{
const NativeProximityParam & param = _params.vector[fieldId];
TermDistanceCalculator::Result result;
const QueryTerm & a = pair.first;
const QueryTerm & b = pair.second;
TermDistanceCalculator::run(a, b, *_md, docId, result);
uint32_t forwardIdx = result.forwardDist > 0 ? result.forwardDist - 1 : 0;
uint32_t reverseIdx = result.reverseDist > 0 ? result.reverseDist - 1 : 0;
feature_t forwardScore = param.proximityTable->get(forwardIdx) * param.proximityImportance;
feature_t reverseScore = param.revProximityTable->get(reverseIdx) * (1 - param.proximityImportance);
feature_t termPairWeight = pair.connectedness *
(a.significance() * a.termData()->getWeight().percent() +
b.significance() * b.termData()->getWeight().percent());
feature_t score = (forwardScore + reverseScore) * termPairWeight / param.maxTableSum;
return score;
}
NativeProximityExecutor::NativeProximityExecutor(const IQueryEnvironment & env,
const NativeProximityParams & params) :
FeatureExecutor(),
_params(params),
_setups(),
_totalFieldWeight(0),
_md(nullptr)
{
std::map<uint32_t, QueryTermVector> fields;
for (uint32_t i = 0; i < env.getNumTerms(); ++i) {
QueryTerm qt = QueryTermFactory::create(env, i);
typedef search::fef::ITermFieldRangeAdapter FRA;
for (FRA iter(*qt.termData()); iter.valid(); iter.next()) {
uint32_t fieldId = iter.get().getFieldId();
if (_params.considerField(fieldId)) { // only consider fields with contribution
qt.fieldHandle(iter.get().getHandle());
fields[fieldId].push_back(qt);
}
}
}
for (std::map<uint32_t, QueryTermVector>::const_iterator itr = fields.begin(); itr != fields.end(); ++itr) {
if (itr->second.size() >= 2) {
FieldSetup setup(itr->first);
generateTermPairs(env, itr->second, _params.slidingWindow, setup);
if (!setup.pairs.empty()) {
_setups.push_back(setup);
_totalFieldWeight += params.vector[itr->first].fieldWeight;
}
}
}
}
void
NativeProximityExecutor::execute(uint32_t docId)
{
feature_t score = 0;
for (size_t i = 0; i < _setups.size(); ++i) {
score += calculateScoreForField(_setups[i], docId);
}
if (_totalFieldWeight > 0) {
score /= _totalFieldWeight;
}
outputs().set_number(0, score);
}
void
NativeProximityExecutor::handle_bind_match_data(const fef::MatchData &md)
{
_md = &md;
}
void
NativeProximityExecutor::generateTermPairs(const IQueryEnvironment & env, const QueryTermVector & terms,
uint32_t slidingWindow, FieldSetup & setup)
{
TermPairVector & pairs = setup.pairs;
for (size_t i = 0; i < terms.size(); ++i) {
for (size_t j = i + 1; (j < i + slidingWindow) && (j < terms.size()); ++j) {
feature_t connectedness = 1;
for (size_t k = j; k > i; --k) {
connectedness = std::min(util::lookupConnectedness(env, terms[k].termData()->getUniqueId(),
terms[k-1].termData()->getUniqueId(), 0.1),
connectedness);
}
connectedness /= (j - i);
if (terms[i].termData()->getWeight().percent() != 0 ||
terms[j].termData()->getWeight().percent() != 0)
{ // only consider term pairs with contribution
pairs.push_back(TermPair(terms[i], terms[j], connectedness));
setup.divisor += (terms[i].significance() * terms[i].termData()->getWeight().percent() +
terms[j].significance() * terms[j].termData()->getWeight().percent()) * connectedness;
}
}
}
}
NativeProximityBlueprint::NativeProximityBlueprint() :
Blueprint("nativeProximity"),
_params(),
_defaultProximityBoost("expdecay(500,3)"),
_defaultRevProximityBoost("expdecay(400,3)")
{
}
NativeProximityBlueprint::~NativeProximityBlueprint()
{
}
void
NativeProximityBlueprint::visitDumpFeatures(const IIndexEnvironment & env,
IDumpFeatureVisitor & visitor) const
{
(void) env;
visitor.visitDumpFeature(getBaseName());
}
Blueprint::UP
NativeProximityBlueprint::createInstance() const
{
return Blueprint::UP(new NativeProximityBlueprint());
}
bool
NativeProximityBlueprint::setup(const IIndexEnvironment & env,
const ParameterList & params)
{
_params.resize(env.getNumFields());
_params.slidingWindow = util::strToNum<uint32_t>(env.getProperties().lookup(getBaseName(), "slidingWindowSize").get("4"));
FieldWrapper fields(env, params, FieldType::INDEX);
vespalib::string defaultProximityImportance = env.getProperties().lookup(getBaseName(), "proximityImportance").get("0.5");
for (uint32_t i = 0; i < fields.getNumFields(); ++i) {
const FieldInfo * info = fields.getField(i);
uint32_t fieldId = info->id();
NativeProximityParam & param = _params.vector[fieldId];
param.field = true;
if ((param.proximityTable =
util::lookupTable(env, getBaseName(), "proximityTable", info->name(), _defaultProximityBoost)) == NULL)
{
return false;
}
if ((param.revProximityTable =
util::lookupTable(env, getBaseName(), "reverseProximityTable", info->name(), _defaultRevProximityBoost)) == NULL)
{
return false;
}
param.fieldWeight = indexproperties::FieldWeight::lookup(env.getProperties(), info->name());
if (param.fieldWeight == 0 ||
info->isFilter())
{
param.field = false;
}
param.proximityImportance = util::strToNum<feature_t>
(env.getProperties().lookup(getBaseName(), "proximityImportance", info->name()).
get(defaultProximityImportance));
if (NativeRankBlueprint::useTableNormalization(env)) {
const Table * fp = param.proximityTable;
const Table * rp = param.revProximityTable;
if (fp != NULL && rp != NULL) {
double value = (fp->max() * param.proximityImportance) +
(rp->max() * (1 - param.proximityImportance));
_params.setMaxTableSums(fieldId, value);
}
}
if (param.field) {
env.hintFieldAccess(fieldId);
}
}
describeOutput("score", "The native proximity score");
return true;
}
FeatureExecutor &
NativeProximityBlueprint::createExecutor(const IQueryEnvironment &env, vespalib::Stash &stash) const
{
NativeProximityExecutor &native = stash.create<NativeProximityExecutor>(env, _params);
if (native.empty()) {
return stash.create<ValueExecutor>(std::vector<feature_t>(1, 0.0));
} else {
return native;
}
}
} // namespace features
} // namespace search
|
