3 Comments

1. 

   Rupert Fiasco says:

   Ask someone on the Lucene mailing list, or maybe on the Acts As Solr mailing list (if you’re acts_as_solr)

   http://groups.google.com/group/acts_as_solr

   January 30, 2008 at 6:03 pm

2. 

   Otis Gospodnetic says:

   Not sure what you mean by exponential. Some examples would help.
   If you need help, either consult Lucene ML or http://sematext.com/

   January 31, 2008 at 4:01 am

3. 

   Mike Grafton says:

   I’ll give an example below, but for the record we solved this by upgrading our version of Lucene (we’re now using 2.1, which is what SOLR 1.2 uses).

   Here’s a little table the shows the relationship we saw between the document boost and the ‘fieldNorm’ component of the score contributed by one of the query terms that hit:

   boost  fieldNorm
   1.0    0.3125    =  (5/16, 2^-1.678)
   2.0    20.0      =  (2^4.3219, 2^1 * 10)
   3.0    256.0     =  (2^8)
   4.0    1280.0    =  (2^10.3219, 2^7 * 10)
   5.0    5120.0    =  (2^12.3219, 2^9 * 10)
   6.0    16384.0   =  (2^14.0)
   7.0    40960.0   =  (2^15.3219, 2^12 * 10)
   8.0    81920.0   =  (2^16.3219, 2^13 * 10)
   10.0   327680.0  =  (2^18.3219, 2^15 * 10)
   

   With the newer code we see a linear relationship.

   January 31, 2008 at 5:55 pm