[Xapian-discuss] Similarity Measures

[Gavin Mendel-Gleason]

Hi, 

I'm currently trying to implement a similarity measure
for xapian.  Ideally I'd like to be able to calculate
the following: 

for document i, and document j

s_ij =  a_ij / ( L_i + L_j + a_ij) 

Where L_i is the number of terms in the document i
and a_ij is:

a_ij = Sum[ t_ik t_jk ] 

Where t_ij is 1 if term "j" occurs in document i.

>From looking at the source code for weights it appears
that the sum should be cut up into peices that can be
calculated incrementally.  Is it possible to calculate
this value within the current weight framework. 



__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com