lemur::retrieval::XLingRetMethod Class Reference

Cross lingual retrieval method. More...

#include <XLingRetMethod.hpp>

Inheritance diagram for lemur::retrieval::XLingRetMethod:

List of all members.


Public Member Functions
	XLingRetMethod (const lemur::api::Index &dbIndex, const lemur::api::Index &background, lemur::dictionary::PDict &dict, lemur::api::ScoreAccumulator &accumulator, double l, double b, bool cacheDR, string &sBM, string &tBM, const lemur::api::Stopper stp=NULL, lemur::api::Stemmer stm=NULL)
	Constructor.
virtual	~XLingRetMethod ()
	clean up.
virtual lemur::api::DocumentRep *	computeDocRep (lemur::api::DOCID_T docID)
	Create a document representation.
virtual double	matchedTermWeight (lemur::api::TERMID_T id, double weight, const lemur::api::DocInfo info, const lemur::api::DocumentRep dRep) const
	Score a given term for a given document.
virtual double	adjustedScore (double origScore, double pge) const
	Adjust the score for a given document.
virtual void	scoreCollection (const lemur::api::QueryRep &qry, lemur::api::IndexedRealVector &results)
virtual void	scoreInvertedIndex (const lemur::api::QueryRep &qryRep, lemur::api::IndexedRealVector &scores, bool scoreAll=false)
virtual lemur::api::QueryRep *	computeQueryRep (const lemur::api::Query &qry)
virtual lemur::api::QueryRep *	computeTargetKLRep (const lemur::api::QueryRep *qry)
virtual double	scoreDoc (const lemur::api::QueryRep &qry, lemur::api::DOCID_T docID)
	Score a document identified by the id w.r.t. a query rep.
virtual void	updateQuery (lemur::api::QueryRep &qryRep, const lemur::api::DocIDSet &relDocs)
	update the query -- noop
Protected Member Functions
virtual double	scoreDocVector (const XLingQueryModel &qRep, lemur::api::DOCID_T docID, lemur::utility::FreqVector &docVector)
Protected Attributes
double	lambda
double	beta
double	numSource
double	numTarget
bool	docBasedSourceSmooth
bool	docBasedTargetSmooth
lemur::api::ScoreAccumulator &	scAcc
lemur::dictionary::PDict &	dictionary
lemur::api::Stemmer *	stemmer
const lemur::api::Stopper *	stopper
const lemur::api::Index &	source
lemur::api::DocumentRep **	docReps
	cache document reps.
bool	cacheDocReps
	whether or not to cache document representations
int	docRepsSize
	number of documents plus 1, the size of the docReps array.
lemur::api::ScoreAccumulator *	termScores

Detailed Description

Cross lingual retrieval method.

Translation dictionary based retrieval, scoring queries in the source language against documents in the target language using:
P(Q_s|D_t) = Prod_w_in_Q_s(lambda(Sum_t_in_D_t P(t|D_t)P(w|t) + (1-lambda)P(w|G_s)
where G_s is the background model for the source language.

Constructor & Destructor Documentation

lemur::retrieval::XLingRetMethod::XLingRetMethod ( const lemur::api::Index & dbIndex,

const lemur::api::Index & background,

lemur::dictionary::PDict & dict,

lemur::api::ScoreAccumulator & accumulator,

double l,

double b,

bool cacheDR,

string & sBM,

string & tBM,

const lemur::api::Stopper * stp = NULL,

lemur::api::Stemmer * stm = NULL

)

Constructor.

Parameters:

dbIndex index for target language documents

background index for source language background model

dict PDict containing source->target translation probabilities

accumulator ScoreAccumulator for intermediate results

l lambda value to use for smoothing background model

b beta value to use for smoothing P(t|D)

cacheDR whether or not to cache document reps

sBM whether to use term frequency (tf/|V|) or term doc frequency (docCount(t)/Sum_w_in_V(docCount(w))) for the source language background model. Default is term frequency.

tBM whether to use term frequency (tf/|V|) or term doc frequency (docCount(t)/Sum_w_in_V(docCount(w))) for the targetlanguage background model. Default is term frequency.

stp source language Stopper to use when getting translations.

stm source language Stemmer to use when getting translations.

lemur::retrieval::XLingRetMethod::~XLingRetMethod ( ) [virtual]

clean up.

Member Function Documentation

virtual double lemur::retrieval::XLingRetMethod::adjustedScore ( double origScore,

double pge

) const [inline, virtual]

Adjust the score for a given document.

Parameters:

origScore the original score

pge the background probability to adjust by.

Returns:
log((lambda * origScore) + ((1 - lambda) * pge))

lemur::api::DocumentRep * lemur::retrieval::XLingRetMethod::computeDocRep ( lemur::api::DOCID_T docID ) [virtual]

Create a document representation.

Parameters:

docID the internal document id to create the representation for

Returns:
An instance of XLingDocRep

virtual lemur::api::QueryRep* lemur::retrieval::XLingRetMethod::computeQueryRep ( const lemur::api::Query & qry ) [inline, virtual]

virtual lemur::api::QueryRep* lemur::retrieval::XLingRetMethod::computeTargetKLRep ( const lemur::api::QueryRep * qry ) [virtual]

virtual double lemur::retrieval::XLingRetMethod::matchedTermWeight ( lemur::api::TERMID_T id,

double weight,

const lemur::api::DocInfo * info,

const lemur::api::DocumentRep * dRep

) const [inline, virtual]

Score a given term for a given document.

Parameters:

id the term id

weight the weight for this term

info the DocInfo for this document

dRep the DocumentRep for this document

Returns:
P(t|D) * P(s|t)

virtual void lemur::retrieval::XLingRetMethod::scoreCollection ( const lemur::api::QueryRep & qry,

lemur::api::IndexedRealVector & results

) [inline, virtual]

virtual double lemur::retrieval::XLingRetMethod::scoreDoc ( const lemur::api::QueryRep & qry,

lemur::api::DOCID_T docID

) [virtual]

Score a document identified by the id w.r.t. a query rep.

double lemur::retrieval::XLingRetMethod::scoreDocVector ( const XLingQueryModel & qRep,

lemur::api::DOCID_T docID,

lemur::utility::FreqVector & docVector

) [protected, virtual]

virtual void lemur::retrieval::XLingRetMethod::scoreInvertedIndex ( const lemur::api::QueryRep & qryRep,

lemur::api::IndexedRealVector & scores,

bool scoreAll = false

) [virtual]

virtual void lemur::retrieval::XLingRetMethod::updateQuery ( lemur::api::QueryRep & qryRep,

const lemur::api::DocIDSet & relDocs

) [inline, virtual]

update the query -- noop