Switching to Adaptive Scoring

Switching to Adaptive Scoring

[Lance A. Brown]

I've been using Gnus 5 since about (ding) version 0.30 and am
considering switching to adaptive scoring.  What is the best way to do
this given that I have a large number of *.SCORE files with various
tweaks in them.

I'm also interested in seeing some examples of adaptive scoring setups
that work well for people.

Thanks,
  --[Lance]

Re: Switching to Adaptive Scoring

[Jason L Tibbitts III]

>>>>> "LAB" == Lance A Brown <labrown@dg-rtp.dg.com> writes:

LAB> I've been using Gnus 5 since about (ding) version 0.30 and am
LAB> considering switching to adaptive scoring.  What is the best way to do
LAB> this given that I have a large number of *.SCORE files with various
LAB> tweaks in them.

You can just turn it on.  All adaptive scoring info by default goes into
*.ADAPT files.  Any adjustments made in your SCORE files still work.

LAB> I'm also interested in seeing some examples of adaptive scoring setups
LAB> that work well for people.

I've tweaked this quite a bit over the last ten months or so.  At one point
it may have been made the default.  The idea is that you use "d" if you
don't like the author and "k" if you don't like the subject.  You might
want to adjust the bonuses and penalties if you read on average a different
percentage of the articles in a group (I average about 2/5) or if you find
author more important than subject.

(setq
 gnus-default-adaptive-score-alist  '((gnus-kill-file-mark)
				     (gnus-unread-mark)
				     (gnus-read-mark
				      (from  3) (subject  30))
				     (gnus-catchup-mark
				      (subject -10))
				     (gnus-killed-mark
				      (from -1) (subject -20))
				     (gnus-expirable-mark
				      (from -2) (subject -15))
				     (gnus-del-mark
				      (from -2) (subject -15))
				     )

Re: Switching to Adaptive Scoring

[Jack Vinson]

Hasn't anyone else found that adaptive scoring does not work properly when
scoring subjects?  I have found that the adaptive scoring mechanism does
not assign the correct score to subjects that it has seen before.  I
suspect that this is the fuzzy matching routine, but I am not sure.

As a result, the only useful adaptive scoring I get is on authors.

The strangest thing is that if I ask for a score trace (V t) on an article
that should have been adaptively scored on its subject, I get the proper
entry from the .ADAPT file, even though the score of the article (V S) does
not jive with the trace.

Anyone?

-- 
Jonathan "Jack" Vinson       jvinson@cheux.ecs.umass.edu
Sunderland, MA               <http://www.cis.upenn.edu/~vinson/home.html>
"Churchill was a shopping bag" - Fatima Mansions

Re: Switching to Adaptive Scoring

[Jason L Tibbitts III]

>>>>> "JV" == Jack Vinson <jvinson@cheux.ecs.umass.edu> writes:

JV> Hasn't anyone else found that adaptive scoring does not work properly
JV> when scoring subjects?

I'm certain that it used to work, but now that I check I find it's not
working.  I'm using sgnus 0.22.

JV> Anyone?

Yep.  The score trace shows that an ADAPT entry is made for the subject but
that it doesn't have any effect on the score.

Oddly, I know that an entry in, say, all.SCORE makes some difference but
all of those entries are exact matches while all of my ADAPT entries are
fuzzy matches.

 - J<

Re: Switching to Adaptive Scoring

[Jack Vinson]

>>>>> "JLT" == Jason L Tibbitts <tibbs@sina.hpc.uh.edu> writes:

>>>>> "JV" == Jack Vinson <jvinson@cheux.ecs.umass.edu> writes:
JV> Hasn't anyone else found that adaptive scoring does not work properly
JV> when scoring subjects?

JLT> Yep.  The score trace shows that an ADAPT entry is made for the
JLT> subject but that it doesn't have any effect on the score.

Even more interesting.  If you manage to get an adaptive entry and then
come back to the group and score the original article as if it had not been
read, then it will be correctly scored.  It's children will not.  This has
me convinced that the algorithm which strips the "Re: " from subject
strings is not doing exactly what we would like.

-- 
Jack Vinson                                                  Sunderland, MA
jvinson@cheux.ecs.umass.edu    <http://www.cis.upenn.edu/~vinson/home.html>
"There's balogna in our slacks."