de.unidu.is.retrieval.pire
Class PIRE

java.lang.Object
  de.unidu.is.retrieval.pire.PIRE

public class PIRE

extends java.lang.Object

An IR engine based on probabilistic Datalog ("Probabilistic Datalog IR Engine").

This IR engine uses a document models consisting of attributes (which refer to a datatype). Datatypes are modelled by special classes in the package de.unidu.is.retrieval.pire.dt, where the class name is the datatype name plus DT.

Probabilistic Datalog is used in an Index. Currently a PDatalogIndex is used, which used the code in de.unidu.is.pdatalog. If another pDatalog implementation should be used, only the method newIndex(String) has to be overwritten in a subclass.

Since:

2003-08-16

Version:

$Revision: 1.30 $, $Date: 2005/03/18 22:02:33 $

Author:

Henrik Nottelmann

Field Summary

protected RelationBase	base The relation base.
protected java.lang.String	collectionName The name of this collection.
protected java.util.Map	conditions Map with temporary conditions for retrieval with weighted-sum queries.
protected PropertyMap	counts Property map with temporary counter for sub-queries.
protected Index	dummyIndex Dummy index used for retrieval.
protected java.util.Map	indexes Indexes, specified by their name.
protected java.util.Map	rules Map with temporary rules for retrieval with Boolean-style queries.
protected Schema	schema Schema.

Constructor Summary

PIRE(DB db, java.lang.String collectionName)
Creates a new instance.

Method Summary

void	addCondition(java.lang.String queryID, java.lang.String attName, java.lang.String operator, double weight, java.lang.Object value) Adds a condition for a weighted sum query.
void	addCondition(java.lang.String queryID, WeightedQueryCondition cond) Adds a condition for a weighted sum query.
void	addConjunction(java.lang.String queryID, QueryCondition[] conditions) Adds a conjunction for a Boolean-style query in disjunctive form.
void	addMomentsCondition(java.lang.String queryID, java.lang.String attName, java.lang.String operator, double weight, java.lang.Object value) Adds a query condition.
void	addMomentsCondition(java.lang.String queryID, WeightedQueryCondition cond) Adds a query condition.
void	addToIndex(java.lang.String docID) Add the document id to the corresponding table.
void	addToIndex(java.lang.String docID, java.lang.String attName, java.lang.Object value) Add the document value of the specified attribute to the index.
void	closeQuery(java.lang.String queryID) Finishes the processing of this query and frees used resources.
void	computeIndex() Computes the index, based on the document values added.
void	computeMoments() Computes the moments of the indexing weights.
void	computeProbs(java.lang.String queryID) Computes probabilities of relevance based on the documents' RSV.
java.lang.String	getCollectionName() Return the name of this collection.
protected java.util.List	getConditionList(java.lang.String queryID) Returns a list of conditions for the specified query id.
protected DT	getDT(SchemaElement element) Returns a data type object for the specified schema element
protected SchemaElement	getElement(java.lang.String attName) Returns the attribute with the specified name.
protected Index	getIndex(java.lang.String key) Returns the index specified by its name.
Index	getIndex(java.lang.String attName, java.lang.String operator) Returns the index specified by its name (created based on the attribute name and the operator name).
Moments	getMoments(java.lang.String queryID) Returns the expectation and the variance of the RSVs.
double	getRD(java.lang.String attName, java.lang.String operator, java.lang.String key) Returns the value corresponding to the specified key in the rd relation in the specified index.
java.util.List	getResult(java.lang.String queryID, int numDocs) Returns the probabilities of relevance for the top-ranked documents in decreasing order.
protected java.util.List	getRuleList(java.lang.String queryID) Returns a list of rules for the specified query id.
void	initIndex() Inits the index.
void	initQuery(java.lang.String queryID) Inits the query.
protected Index	newIndex(java.lang.String key) Creates a new PDatalog index with the specified name.
void	registerAttribute(java.lang.String attName, java.lang.String datatype, java.util.List operators) Registers the specified attribute.
void	removeIndex() Removes the index.
void	setRD(java.lang.String attName, java.lang.String operator, java.lang.String key, double value) Sets the value corresponding to the specified key in the rd relation in the specified index.

Methods inherited from class java.lang.Object

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

Field Detail

collectionName

protected java.lang.String collectionName

The name of this collection.

indexes

protected java.util.Map indexes

Indexes, specified by their name.

schema

protected Schema schema

Schema.

conditions

protected java.util.Map conditions

Map with temporary conditions for retrieval with weighted-sum queries. Keys are query IDs (as strings), values are lists of rules (as WeightedQueryCondition objects).

rules

protected java.util.Map rules

Map with temporary rules for retrieval with Boolean-style queries. Keys are query IDs (as strings), values are lists of rules (as Rule objects).

counts

protected PropertyMap counts

Property map with temporary counter for sub-queries. Keys are query IDs (as strings), values are the counters (as integers).

dummyIndex

protected Index dummyIndex

Dummy index used for retrieval.

base

protected RelationBase base

The relation base.

Constructor Detail

PIRE

public PIRE(DB db,
            java.lang.String collectionName)

Creates a new instance.

Parameters:

db - database parameters

collectionName - name of the collection

Method Detail

getDT

protected DT getDT(SchemaElement element)

Returns a data type object for the specified schema element

Parameters:

element - schema element

Returns:

data type object

getIndex

public Index getIndex(java.lang.String attName,
                      java.lang.String operator)

Returns the index specified by its name (created based on the attribute name and the operator name). If no index exists so far, a new one will be created a added to the indexes map.

Parameters:

attName - attribute name

operator - name

Returns:

corresponding index

getIndex

protected Index getIndex(java.lang.String key)

Returns the index specified by its name. If no index exists so far, a new one will be created a added to the indexes map.

Parameters:

key - index name

Returns:

corresponding index

newIndex

protected Index newIndex(java.lang.String key)

Creates a new PDatalog index with the specified name.

Parameters:

key - index name

Returns:

new, empty index

getElement

protected SchemaElement getElement(java.lang.String attName)

Returns the attribute with the specified name.

Parameters:

attName - attribute name

Returns:

attribute definition

registerAttribute

public void registerAttribute(java.lang.String attName,
                              java.lang.String datatype,
                              java.util.List operators)

Registers the specified attribute. This only has to be called once per attribute after creating this IR engine.

Parameters:

attName - attribute name

datatype - corresponding datatype

operators - list of operators

getRD

public double getRD(java.lang.String attName,
                    java.lang.String operator,
                    java.lang.String key)

Returns the value corresponding to the specified key in the rd relation in the specified index.

Parameters:

attName - schema attribute name

operator - search operator

key - value key

Returns:

value

setRD

public void setRD(java.lang.String attName,
                  java.lang.String operator,
                  java.lang.String key,
                  double value)

Sets the value corresponding to the specified key in the rd relation in the specified index.

Parameters:

attName - schema attribute name

operator - search operator

key - value key

value - value

initIndex

public void initIndex()

Inits the index.

addToIndex

public void addToIndex(java.lang.String docID)

Add the document id to the corresponding table.

Parameters:

docID - document id

addToIndex

public void addToIndex(java.lang.String docID,
                       java.lang.String attName,
                       java.lang.Object value)

Add the document value of the specified attribute to the index. Parsing the value is left to the datatype implementation; so, for text this can be a fulltext string.

Parameters:

docID - document id

attName - attribute name

value - attribute value

computeIndex

public void computeIndex()

Computes the index, based on the document values added.

computeMoments

public void computeMoments()

Computes the moments of the indexing weights.

removeIndex

public void removeIndex()

Removes the index.

initQuery

public void initQuery(java.lang.String queryID)

Inits the query.

Parameters:

queryID - query id

addCondition

public void addCondition(java.lang.String queryID,
                         java.lang.String attName,
                         java.lang.String operator,
                         double weight,
                         java.lang.Object value)

Adds a condition for a weighted sum query. These conditions are collected and evaluated later in computeProbs(String), as conditions for the same attribute/operator pair have to be evaluated together.

Parameters:

queryID - query id

attName - schema attribute name

operator - search operator

weight - condition weight

value - comparison value

addCondition

public void addCondition(java.lang.String queryID,
                         WeightedQueryCondition cond)

Adds a condition for a weighted sum query. These conditions are collected and evaluated later in computeProbs(String), as conditions for the same attribute/operator pair have to be evaluated together.

Parameters:

queryID - query id

cond - weighted query condition

addConjunction

public void addConjunction(java.lang.String queryID,
                           QueryCondition[] conditions)

Adds a conjunction for a Boolean-style query in disjunctive form. The conjunction forms one term in the disjunctive form. This method computes both the RSV and the probabilities of relevance for the given conditions, and adds a rule for computing the conjunction (of the probabilities of relevance) to the dummy index. These rules are later evaluated together in computeProbs(String).

Parameters:

queryID - query id

conditions - conditions forming a conjunction

computeProbs

public void computeProbs(java.lang.String queryID)

Computes probabilities of relevance based on the documents' RSV.

Parameters:

queryID - query id

getResult

public java.util.List getResult(java.lang.String queryID,
                                int numDocs)

Returns the probabilities of relevance for the top-ranked documents in decreasing order. Before this, computeProbs() has to be called.

Parameters:

queryID - query id

numDocs - number of documents to retrieve

Returns:

list of ProbDoc instances

closeQuery

public void closeQuery(java.lang.String queryID)

Finishes the processing of this query and frees used resources.

Parameters:

queryID - query id

addMomentsCondition

public void addMomentsCondition(java.lang.String queryID,
                                java.lang.String attName,
                                java.lang.String operator,
                                double weight,
                                java.lang.Object value)

Adds a query condition. The results for that condition are inserted into the results (for expectation and variance).

Parameters:

queryID - query id

attName - schema attribute name

operator - search operator

weight - condition weight

value - comparison value

addMomentsCondition

public void addMomentsCondition(java.lang.String queryID,
                                WeightedQueryCondition cond)

Adds a query condition. The results for that condition are inserted into the results (for expectation and variance).

Parameters:

queryID - query id

cond - weighted query condition

getMoments

public Moments getMoments(java.lang.String queryID)

Returns the expectation and the variance of the RSVs.

Parameters:

queryID - query id

Returns:

moments

getCollectionName

public java.lang.String getCollectionName()

Return the name of this collection.

Returns:

name of this collection

getRuleList

protected java.util.List getRuleList(java.lang.String queryID)

Returns a list of rules for the specified query id.

Parameters:

queryID - query id

Returns:

list of temporary rules

getConditionList

protected java.util.List getConditionList(java.lang.String queryID)

Returns a list of conditions for the specified query id.

Parameters:

queryID - query id

Returns:

list of temporary conditions