in reply to Related module links on search.cpan.org

I've seen this recently on CPAN; it's actually kind of neat and actually pretty helpful overall. Though, as you noted:

...information is automated, relating the appearance of modules referenced in the same article. This automation is good because it is automatically maintained, but it is dull sometimes relating, for example, LWP::UserAgent and Data::Dumper...

It is a lot like the Amazon.com strategy of the Other people who looked at this item also looked at.... And although the CPAN version just says Related Modules, I have generally just used in the spirit of the Amazon.com strategy.

There is certainly room for improvement, IMHO, but is still a nice addition to CPAN.

ack Albuquerque, NM
  • Comment on Re: Related module links on search.cpan.org

Replies are listed 'Best First'.
Re^2: Related module links on search.cpan.org (stats)
by tye (Sage) on Apr 11, 2008 at 18:01 UTC

    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. :)

    - tye        

[reply]

      In a nutshell:

      • We have a list of matching modules sorted by "tf"
      • We need a list of matching modules sorted by the weighted "tf/idf"
[reply]

In Section Perl Monks Discussion