-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello folks,

we want to work with explanations of document scores inside result lists.
In this context we are interested on the scores of the single terms from a
query, for each document inside the result list:

Query:
"termA termB"

Result:
doc1  => overall score 2.3
doc2  => overall score 1.6

We would love it to have explanations like this:
overall score doc1 (2.3) = score termA (1.2) + score termB (1.1)
overall score doc2 (1.6) = score termA (1.1) + score termB (0.5)

In the past we worked with the Searcher.explain(..) method which is appropriate
in order to explain the results for small numbers of documents, but since this
method takes as much time as a whole search (as written inside the ApiDoc), this
of course is not feasible for whole result lists.

Nevertheless, all values should be available during the calculation of the 
overall
score, which is done inside the Similarity class. Thus, collecting of these 
should
result into nearly no runtime overhead, its mainly a question about memory.

We have looked inside Similarity, and all is available except the current 
document
id - so we have term score values but we don't know the documents they are 
related
to. And this is our question:
Does anybody know how to get this current document number/id inside a subclass
implementation of Similarity?


Thanks in advance!

Chris




- --
______________________________________________________________________

Christian Reuschling, Dipl.-Ing.(BA)
Software Engineer

Knowledge Management Department
German Research Center for Artificial Intelligence DFKI GmbH
Erwin-Schrödinger-Straße 57, D-67663 Kaiserslautern, Germany

Phone: +49.631.205-3441
mailto:[EMAIL PROTECTED]  http://www.dfki.uni-kl.de/~reuschling/
______________________________________________________________________
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2 (GNU/Linux)

iD8DBQFFVGclQoTr50f1tpcRArjFAJ44DUfNDCdjhj3H/nsh7lWQD4nb7QCgmnXM
SHA8byhpGUZJyA/mwQ7IIws=
=Uka/
-----END PGP SIGNATURE-----

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to