The ATIRE verison of BM25. More...

Public Member Functions
	ranking_function_atire_bm25 (double k1, double b, std::vector< compress_integer::integer > &document_lengths)
	Constructor. More...

	~ranking_function_atire_bm25 ()
	Destructor.

forceinline void	compute_idf_component (compress_integer::integer document_frequency, compress_integer::integer documents_in_collection)
	Called once per term (per query). Computes the IDF component of the ranking function and stores it internally. More...

forceinline void	compute_tf_component (index_postings_impact::impact_type term_frequency)
	Compute and store internally the term-frequency based component of the ranking function (useful when postings lists are impact ordered) More...

forceinline double	compute_score (compress_integer::integer document_id, index_postings_impact::impact_type term_frequency)
	Compute BM25 from the given document, assuming pieces have already been computed. More...

Static Public Member Functions
static void	unittest (void)
	Unit test this class.

Private Attributes
double	idf
	the IDF of the term being processed

double	top_row
	the top-row of the ranking function for the term being processed (tf(td) * (k1 + 1))

double	k1_plus_1
	k1 + 1

double	mean_document_length
	the mean of the document lengths

std::vector< float >	length_correction
	most of the bottom row of BM25 (k1 * ((1 - b) + b * length / mean_document_length)) for the current term being processed

Detailed Description

The ATIRE verison of BM25.

Constructor & Destructor Documentation

JASS::ranking_function_atire_bm25::ranking_function_atire_bm25	(	double	k1,
		double	b,
		std::vector< compress_integer::integer > &	document_lengths
	)

inline

Constructor.

Parameters

k1	[in] the BM25 k1 parameter, 0.9 is a good value.
b	[in] the BM25 b parameter, 0.4 is a good value.
document_lengths	[in] a vector holding the length of each document in the collection.

forceinline void JASS::ranking_function_atire_bm25::compute_idf_component	(	compress_integer::integer	document_frequency,
		compress_integer::integer	documents_in_collection
	)

inline

Called once per term (per query). Computes the IDF component of the ranking function and stores it internally.

Parameters

document_frequency	[in] the number of documents that contain this term.
documents_in_collection	[in] the number of documents in the collection.

forceinline double JASS::ranking_function_atire_bm25::compute_score	(	compress_integer::integer	document_id,
		index_postings_impact::impact_type	term_frequency
	)

inline

Compute BM25 from the given document, assuming pieces have already been computed.

First compute the IDF for the term using compute_idf_component(), then the TF component using compute_tf_component(), then call this.

Parameters

document_id	[in] The ID of the document (used to look up the length)
term_frequency	[in] The number of times the term occurs in the document.

forceinline void JASS::ranking_function_atire_bm25::compute_tf_component ( index_postings_impact::impact_type term_frequency )

inline

Compute and store internally the term-frequency based component of the ranking function (useful when postings lists are impact ordered)

Parameters

term_frequency [in] The number of times the term occurs in the document.

The documentation for this class was generated from the following file: