Yes, it would be helpful to normalize the scores so that Data::Dumper isn't the module considered most closely "related" to just about every module on CPAN just because it is the most mentioned on PerlMonks (and so is the "most mentioned in conjuction with X" for most values of "X").

So it'd be cool to be able to list modules "related to X" sorted by how much more likely they are to be mentioned when "X" is mentioned. I think even the math for that is quite simple. Then Data::Dumper could even have a negative score in relation to some module. :)