Ideas to improve scoring

Hash: SHA1

Ok, I am getting more familiar with C# (as some of you know I am
hoping to become a more productive user) But these little quarks in
scoring have been nagging me, I have played with some simple fixes,
but none of them have worked very well..... regardless, here it goes.

Ok, so if I run a query for Environmental, I would hope to turn up
something that might relate to the environment, the problem is, our
search term isn't broken down. I tried at first to create a simple set
of rules (like easy plurals, if a word ends in 's' run the query
without the 's' as well etc) but there's really nothing universal.
What about harnessing a pre-exisiting spelling lib to offer some of
this functionality? Again, not familiar on the specifics, but offer a
check box in best or beagle-settings that allows for 'fuzzy-searches'
which also query works within a certain lexiconal range.

Something to consider for Holmes (should be decide to integrate a
linking to some spelling lib/program such as gtk-spell or aspell or
ispell or whatever) would be a 'did you mean ____' type prompt, I
think integration into best atm isn't worth the effort, but if were
thinking ground up, why not.

The issue at this point comes with scoring these parallel results,
since we can't live query multiple queries and then sort them on
arrival, we would have to query each word individually, and based on
is lexiconal difference from the original, adjust its scoring,. This
part would no doubt be the hardest, but an alpha implementation
probably wouldn't want to include it.

I dunno, the idea is rought, and I currently have been downed with the
flu, thus heavy doses of NyQuill have inspired this, if its completely
impractial/beyond the scope of this project, then so be it, otherwise,
lets discuss it and try to make a working model.

- --
Kevin Kubasik

Version: GnuPG v1.4.2 (GNU/Linux)


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]