Package org.terrier.querying

Class QueryExpansion

java.lang.Object

org.terrier.querying.QueryExpansion

All Implemented Interfaces:

MQTRewritingProcess, Process

public class QueryExpansion
extends java.lang.Object
implements MQTRewritingProcess

Implements automatic query expansion as PostProcess that is applied to the result set after 1st-time matching. Controls

• qemodel : The query expansion model used for Query Expansion. Defaults to Bo1.
• qe_fb_terms : The maximum number of most weighted terms in the pseudo relevance set to be added to the original query. The system performs a conservative query expansion if this property is set to 0. A conservative query expansion only re-weighs the original query terms without adding new terms to the query. Default set by property expansion.terms.
• qe_fb_docs : The number of top documents from the 1st pass retrieval to use for QE. The query is expanded from this set of documents, also known as the pseudo relevance set. Default set by property expansion.docs.

Properties

qe.feedback.selector : The class to be used for selecting feedback documents. qe.expansion.terms.class : The class to be used for selecting expansion terms from feedback documents.

Author:

Gianni Amatti, Ben He, Vassilis Plachouras, Craig Macdonald

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static class	QueryExpansion.QueryExpansionConfig

Field Summary

Fields

Modifier and Type	Field	Description
protected java.util.Map<java.lang.String,QueryExpansionModel>	Cache_QueryExpansionModel	Caching the query expansion models that have been created so far.
protected CollectionStatistics	collStats	The statistics of the index
protected PostingIndex<?>	directIndex	The direct index used for retrieval.
protected DocumentIndex	documentIndex	The document index used for retrieval.
protected java.lang.String[]	expanderNames
protected PostingIndex<?>	invertedIndex	The inverted index used for retrieval.
java.lang.String	lastExpandedQuery
protected Index	lastIndex
protected Lexicon<java.lang.String>	lexicon	An instance of Lexicon class.
protected static org.slf4j.Logger	logger
protected MetaIndex	metaIndex
static java.lang.String	NAMESPACE_QEMODEL	The default namespace of query expansion model classes.
protected QueryExpansionModel	QEModel	The query expansion model used.
protected FeedbackSelector	selector	The process by which to select feedback documents
protected boolean	warnQeModel	Set to true if this instancae has warned about the control "qemodel" not being set

Constructor Summary

Constructors

Constructor	Description
QueryExpansion()	The default constructor of QueryExpansion.
QueryExpansion(java.lang.String[] expansionTerms)

Method Summary

Modifier and Type	Method	Description
static QueryExpansion.QueryExpansionConfig	configFromRequest(Request r)	Get the standard configuration for QueryExpansion, namely number of terms and docs, from the Request, using two controls.
void	configureIndex(Index index)
boolean	expandQuery(MatchingQueryTerms query, Request rq)	This method implements the functionality of expanding a query.
protected ExpansionTerms	getExpansionTerms()	load the expansion terms, as per the property qe.expansion.terms.class.
protected FeedbackSelector	getFeedbackSelector(Request rq)	load the feedback selector, based on the property qe.feedback.selector
protected Index	getIndex(Manager m)	For easier sub-classing of which index the query expansion comes from
java.lang.String	getInfo()	Returns the name of the used query expansion model.
QueryExpansionModel	getQueryExpansionModel(java.lang.String Name)	Obtain the query expansion model for QE to use.
void	process(Manager manager, Request q)	Runs the actual query expansion
void	setQueryExpansionModel(QueryExpansionModel _QEModel)	Set the used query expansion model.

Methods inherited from class java.lang.Object

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

Methods inherited from interface org.terrier.querying.Process

process

Field Detail

logger

protected static final org.slf4j.Logger logger

NAMESPACE_QEMODEL

public static final java.lang.String NAMESPACE_QEMODEL

The default namespace of query expansion model classes. The query expansion model names are prefixed with this namespace, if they are not fully qualified.

See Also:

Constant Field Values

Cache_QueryExpansionModel

protected java.util.Map<java.lang.String,QueryExpansionModel> Cache_QueryExpansionModel

Caching the query expansion models that have been created so far.

documentIndex

protected DocumentIndex documentIndex

The document index used for retrieval.

metaIndex

protected MetaIndex metaIndex

invertedIndex

protected PostingIndex<?> invertedIndex

The inverted index used for retrieval.

lexicon

protected Lexicon<java.lang.String> lexicon

An instance of Lexicon class.

directIndex

protected PostingIndex<?> directIndex

The direct index used for retrieval.

warnQeModel

protected boolean warnQeModel

Set to true if this instancae has warned about the control "qemodel" not being set

expanderNames

protected java.lang.String[] expanderNames

collStats

protected CollectionStatistics collStats

The statistics of the index

QEModel

protected QueryExpansionModel QEModel

The query expansion model used.

selector

protected FeedbackSelector selector

The process by which to select feedback documents

lastIndex

protected Index lastIndex

lastExpandedQuery

public java.lang.String lastExpandedQuery

Constructor Detail

QueryExpansion

public QueryExpansion()

The default constructor of QueryExpansion.

QueryExpansion

public QueryExpansion(java.lang.String[] expansionTerms)

Method Detail

configFromRequest

public static QueryExpansion.QueryExpansionConfig configFromRequest(Request r)

Get the standard configuration for QueryExpansion, namely number of terms and docs, from the Request, using two controls.

setQueryExpansionModel

public void setQueryExpansionModel(QueryExpansionModel _QEModel)

Set the used query expansion model.

Parameters:

_QEModel - QueryExpansionModel The query expansion model to be used.

expandQuery

public boolean expandQuery(MatchingQueryTerms query,
                           Request rq)
                    throws java.io.IOException

This method implements the functionality of expanding a query.

Specified by:

expandQuery in interface MQTRewritingProcess

Parameters:

query - MatchingQueryTerms the query terms of the original query.

rq - the Request thus far, giving access to the query and the result set

Throws:

java.io.IOException

getIndex

protected Index getIndex(Manager m)

For easier sub-classing of which index the query expansion comes from

getExpansionTerms

protected ExpansionTerms getExpansionTerms()

load the expansion terms, as per the property qe.expansion.terms.class. Defaults to DFRBagExpansionTerms.

Returns:

an ExpansionTerms instance, which may or may not wrap other ExpansionTerms instances

getFeedbackSelector

protected FeedbackSelector getFeedbackSelector(Request rq)

load the feedback selector, based on the property qe.feedback.selector

process

public void process(Manager manager,
                    Request q)

Runs the actual query expansion

Specified by:

process in interface Process

See Also:

Process.process(org.terrier.querying.Manager,org.terrier.querying.SearchRequest)

configureIndex

public void configureIndex(Index index)

Specified by:

configureIndex in interface MQTRewritingProcess

getQueryExpansionModel

public QueryExpansionModel getQueryExpansionModel(java.lang.String Name)

Obtain the query expansion model for QE to use. This will be cached in a hashtable for the lifetime of the application. If Name does not contain ".", then NAMESPACE_QEMODEL will be prefixed to it before loading.

Parameters:

Name - the name of the query expansion model to load.

getInfo

public java.lang.String getInfo()

Returns the name of the used query expansion model.

Specified by:

getInfo in interface Process

Returns:

String the name of the used query expansion model.