Scoring oddities

Scoring oddities

[Lars Magne Ingebrigtsen]

I have gotten several reports over the last, uh, months about scoring
working badly.  More precisely, what is happening is that scoring
doesn't seem to occur for some articles.  A typical scenario is:

1) You enter a group, and some articles that you think should have
scores don't, and `V S' reports "0".

2) `V t' reports one or several matching score rules.

I have not been able to reproduce it.  Could y'all check around a bit
and report back to this list any anomalies and stuff?

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@ifi.uio.no * Lars Magne Ingebrigtsen

Re: Scoring oddities

[Harald Meland]

[Lars Magne Ingebrigtsen]

| I have not been able to reproduce it.  Could y'all check around a bit
| and report back to this list any anomalies and stuff?

Not sure this problem is related, but I had some troubles getting the
"advanced scoring" stuff to work.

Reading the postmaster mail here kinda takes a lot of time with all
the (malformed, and therefore rejected) spam, so I thought I'd score
down the most frequently encountered candidates (as I believe there is
no way of splitting mail into different groups on body content).
After some testing, I arrived at this score file:

((mark -100)
 ((and ("subject" "Postmaster notify: User unknown" E)
       (or ("body" "^>>> MAIL From:<\\([^@]*\\|.*@[^.]+\\)>$" r)
	   ("body" "^>>> MAIL From:.*\\.\\(com\\|edu\\|net\\|org\\|web\\)>$" r)
	   (not ("body" "^>>> MAIL From:" r))))
  -200))

However, the articles that should have been hit, weren't (neither by
"V S" nor "V t").  After adding a dummy rule to the score file, like
this:

((mark -100)
 ((and ("subject" "Postmaster notify: User unknown" E)
       (or ("body" "^>>> MAIL From:<\\([^@]*\\|.*@[^.]+\\)>$" r)
	   ("body" "^>>> MAIL From:.*\\.\\(com\\|edu\\|net\\|org\\|web\\)>$" r)
	   (not ("body" "^>>> MAIL From:" r))))
  -200)
 ("subject" ("Postmaster notify: User unknown" 0 nil E)))

, flushing the score cache ("V F"), quitting the mail group with "Q"
and then re-entering, the articles were scored as they should be (both
according to "V S" and "V t").

I've tested this again while writing this, and it seems like most
times "V F" and "V R" is enough to get the articles rescored with the
updated score file -- however, sometimes "V F", "Q" and re-entering is
needed.  I see no pattern behind this.
-- 
Harald

Re: Scoring oddities

[Matt Armstrong]

Lars Magne Ingebrigtsen <larsi@ifi.uio.no> writes:

> I have gotten several reports over the last, uh, months about
> scoring working badly.  More precisely, what is happening is that
> scoring doesn't seem to occur for some articles.  A typical scenario
> is:
> 
> 1) You enter a group, and some articles that you think should have
> scores don't, and `V S' reports "0".
> 
> 2) `V t' reports one or several matching score rules.
> 
> I have not been able to reproduce it.  Could y'all check around a
> bit and report back to this list any anomalies and stuff?

Well, I got it this this morning.  After reading a few groups that
were properly scored, I went into one a group, read a few messages,
and realized nothing was scored.

There usually are a lot of articles scored because I have turned on
adaptive scoring.  This using the first post-quassia beta (5.6.1?) on
emacs 19.34 Linux.

I dutifully said "hey, lets see if I can reproduce it", quit back to
the group buffer, re-entered the group, and found that the scoring was
done correctly.  :-(


-- 
matta