new scoring ideas.

[Wes Hardaker <wjhardaker@ucdavis.edu>]

Ok, I've been thinking (always a bad thing) about a fairly easy (I
think) thing to do with adaptive scoring (including word scoring)
(aside: can you score score someone down for using too many ()
comments?)

First off, I have the following ideas about scoring:

1) I want the subject line scoring to stay the same..  Down for
   unread, up for read, just like it is now.

2) I want the author scoring to stay the same, minus one exception:  I
   don't want an author's score to ever drop below 0.

This means that the subject is more important to me than the author's
score, and thus should be adjusted down because I specifically haven't
read that subject before.  However, I don't want to nix out an author
entirely just because he writes about a subject that I don't want to
read.  I want it to drop to 0 only, because I want to see new subjects
(ie, with a score of 0) still be listed no lower than 0 in score.  (I
rarely read anything in the italics section at the bottom).  However,
if I still want gnus to decide for me that some author is really
really cool and to score up an author if I've read him/her a lot.

Make sense?  It should.  Sort of what I'm suggesting for:

(setq gnus-adapt-lower-limits
	'(("subject" -100)
	  ("from" 0)
	  -10000)                ;; default  (nil = no limit)
      gnus-adapt-upper-limits
	nil)

Also, this is even more important for word adaption.  I never use it
because all my articles in large volume groups drop too low in score,
if I read only 1% of the articles.  This would fix this problem, as no
word could ever get scored lower than X but could still get raised,
which would raise the articles I'm probably most interested in but not
those that I'm not.

(setq gnus-adapt-word-lower-limits 0)  ;  defaults to nil

Wes
(ideas?)
-- 
"Ninjas aren't dangerous.  They're more afraid of you than you are of them."