viterbi.sql_in File Reference

concatenate a set of input values into arrays to feed into viterbi c function and create a human readable view of the output More...

Functions
text	vcrf_top1_table (text segment_tbl, text label_tbl, text resulttbl_raw, text result_tbl)
	This function creates the specified result_tbl containing top1 labels for the sequence. More...
int []	vcrf_top1_label (float8[] marray, float8[] rarray, int nlabel)
	This function implements the Viterbi algorithm which takes the sentence to be label as input and return the top1 labeling for that sentence. More...
text	vcrf_label (text segment_tbl, text factor_mtbl, text factor_rtbl, text label_tbl, text result_tbl)
	This function prepares the inputs for the c function 'vcrf_top1_label' and invoke the c function. More...

Detailed Description

Date

February 2012

Function Documentation

vcrf_label()

text vcrf_label	(	text	segment_tbl,
		text	factor_mtbl,
		text	factor_rtbl,
		text	label_tbl,
		text	result_tbl
	)

Parameters

segment_tbl	Name of table containing all the testing sentences.
factor_mtbl	Name of table containing all the m factors.
factor_rtbl	Name of table containing all the r factors.
label_tbl	Name of table containing all the labels in the label space.
result_tbl	Name of table to store the output

Returns

the top1 label sequence, the last two elements in the array is used to calculate the top1 probability

vcrf_top1_label()

int [] vcrf_top1_label	(	float8 []	marray,
		float8 []	rarray,
		int	nlabel
	)

Parameters

marray	Name of arrays containing m factors
rarray	Name of arrays containing r factors
nlabel	Total number of labels in the label space

Returns

the top1 label sequence, the last two elements in the array is used to calculate the top1 probability

vcrf_top1_table()

text vcrf_top1_table	(	text	segment_tbl,
		text	label_tbl,
		text	resulttbl_raw,
		text	result_tbl
	)

Parameters

segment_tbl	Name of table containing all the testing sentences.
label_tbl	Name of table containing all the labels in the label space.
result_tbl	Name of table storing the best label sequence and the conditional probability.