a4tunado / google-all-pairs-similarity-search Goto Github PK

Roberto: Sorry for the late reply but for whatever reason, the first
notification about your Jan 2nd question got lost in my spam filter.
Since you closed the original ticket I am opening a new one with
clarifications.

What I meant is the ability to provide as an input not one dataset but two
dataset. 

In this setting, one dataset would be some "reference" and the second
dataset a "query" dataset. 
The goal would be to find all items in the "query" set that are similar to
items in the "reference" data set above a certain threshold: basically
returning the similarity intersection between the two sets as opposed to
the current setting where only pairs within the same are considered. I
guess one way could be to merge the sets and discard pairs returned from
the same set, though that does seem pretty naive.  

The dblp_le.bin from the downloads section uploaded Aug 2007 does not satisfy 
the input requirements of the all-pairs implementation: there are a handful 
of vectors which contain duplicate features. This leads to a few strange 
results.

[deleted issue]

I wonder if you would have some pointer to use the all pair comparison code
between two lists instead of within one list