Re: adapt in SCORE file does not work logicaly

Re: adapt in SCORE file does not work logicaly

[Emanuel Berg]

Stefan Huchler <stefan.huchler@mail.de> writes:

> thx for an answer!

This answer doesn't answer any of your questions, but
you are welcome nonetheless :)

There is a huge section in the Gnus manual on
scoring. Did you check that out?

For what reason do you want scoring?

If it is to block spammers and perhaps (f)lamers in
general, for mail, I managed to do that with
splitting, just split them to a spam group, then give
that group low priority so it won't even show.

For news, I have a KILL file. I just add addresses to
both. But to KILL isn't as good as splitting because
the to-be killed posts still show up in the group
buffer. It'll say "some.group 6" and you think, cool,
some interesting discussion, hit RET, and... turns
out, all six were KILL-ed.

I just feel it would be better to have a common
interface for splitting both news and mails and also
to have the plonkies in a textfile, i.e., not hard
coded.

Now I have the splitting hard-coded in an Elisp (.el)
file and the killing in the KILL file that is sort of
a hybrid, with entries like

(gnus-kill "From" "Joe Flamer" '(gnus-summary-mark-as-read nil "X") t)

I do have an Elisp function and a shell script to
append automatically, but it is still suboptimal from
the scenario I describe - I don't know how other
people do it.

If you are interested in any of this, I'll be happy to
provide exact examples and the code itself - but
probably you know of this, right?

Anyway, as for scoring, I never got that to work but
it would be very interesting to have features like:

* threads where you are active, bring that to your
  attention
  
* people who quote you, even more so

* subjects that has been around for ages and you never
  participated, downgrade

* very short messages that contains +1, downgrade
  (ignore), as they don't bring anything in terms of
  thoughts or technology

* messages that contains insults ("idiot"), ignore
  (even when true, never pleasant to read, also
  doesn't bring anything)

I don't know if or to what extent that can be done
with scoring, but I would be very happy to learn.

At the very least, a common interface to killing
(common to mail and news) and a clear separation of
code and data should be very possible. I'd be
surprised if no one ever did anything to that extent,
if so, tell me and I'll se what I can do whenever I
get, eh, "time" (ha).

Keep Gnusing!
  
-- 
underground experts united

Re: adapt in SCORE file does not work logicaly

[Ray]

Emanuel Berg <embe8573@student.uu.se> writes:

> Stefan Huchler <stefan.huchler@mail.de> writes:
>
>> thx for an answer!
>
> This answer doesn't answer any of your questions, but
> you are welcome nonetheless :)
>
> There is a huge section in the Gnus manual on
> scoring. Did you check that out?
>
> For what reason do you want scoring?
>
> If it is to block spammers and perhaps (f)lamers in
> general, for mail, I managed to do that with
> splitting, just split them to a spam group, then give
> that group low priority so it won't even show.
>
> Anyway, as for scoring, I never got that to work but
> it would be very interesting to have features like:
>
> * threads where you are active, bring that to your
>   attention
>   
> * people who quote you, even more so
>
> * subjects that has been around for ages and you never
>   participated, downgrade
>
> * very short messages that contains +1, downgrade
>   (ignore), as they don't bring anything in terms of
>   thoughts or technology
>
> * messages that contains insults ("idiot"), ignore
>   (even when true, never pleasant to read, also
>   doesn't bring anything)
>
> I don't know if or to what extent that can be done
> with scoring, but I would be very happy to learn.
>
> At the very least, a common interface to killing
> (common to mail and news) and a clear separation of
> code and data should be very possible. I'd be
> surprised if no one ever did anything to that extent,
> if so, tell me and I'll se what I can do whenever I
> get, eh, "time" (ha).
>
> Keep Gnusing!

They are the very same features that I am looking to have in Gnus. 
I shall follow this thread with interest and offer what I can
when I can. 

RayH

Re: adapt in SCORE file does not work logicaly

[Stefan Huchler]

lol

scoring itself is pretty easy, most of the time I use it for youtube rss
"posts" so maybe its a bit differnt for me, but u just press L for lower
then s for subject then s for substring then p for permanent then u
choose the substring u want to lower in subjects or select cut out the
substring enter done.

how much it lowers the score I think there is some default but u can set
it, and u can set when messages get automaticly greyed out or hidden
completly or something like that.

The problem I have is not about scoring, its about ADAPTIVE scoring, so
that it automaticly get lowered or increased score values when u read it
or kill it or catchup to it etc.

and I even understand how that works in theory, I just dont want to use
it for every group but only one.

Hope that helped YOU and that somebody helps ME, now :)

Re: adapt in SCORE file does not work logicaly

[Stefan Huchler]

btw scoring is better than killing and what u do, because u have no 0-1
switch but u can have different levels of output quality filters, as
example hide thhreads that fall into several fitlers and have a really
bad (low) value, u could even sort by value and see the one u are
interested most first and so on.

I am just tired to hit 50 times a day L s s p and L a e p (author exact
permanenct) for every video, when I know it could happen more or less
automaticly or by pressing only k for everthing I dont like.

Re: adapt in SCORE file does not work logicaly

[Emanuel Berg]

Stefan Huchler <stefan.huchler@mail.de> writes:

> btw scoring is better than killing and what u do,
> because u have no 0-1 switch but u can have
> different levels of output quality filters, as
> example hide thhreads that fall into several fitlers
> and have a really bad (low) value, u could even sort
> by value and see the one u are interested most first
> and so on.
>
> I am just tired to hit 50 times a day L s s p and L
> a e p (author exact permanenct) for every video,
> when I know it could happen more or less automaticly
> or by pressing only k for everthing I dont like.

Scoring is perhaps better than killing and what I do,
but this case doesn't show it because I never hit
anything, and certainly not a four-key combination, 50
times a day.

What you describe from scoring is some kind of
interactive scoring where you score what you like (and
dislike) in different ways.

I'm not interested in that kind of scoring, I'm
interested in me manually setting up rules and then
have Gnus doing the scoring based on those rules and
the input stream of messages, and then show me a
"transformation" of that stream, transparently.

Right now the splitting does that, and the killing
does almost that (except it shows the messages as a
digit in the group buffer before it kills them).

Also, splitting and killing are very basic at this
point (for me at least), they just split (to spam) or
kill based on name and subject.

I would like a more intelligent approach where the
contents of the message is assessed rather than just a
couple of static parameters.

I.e, say some guy posts 100 posts, 10 are flame war
posts, 10 are "1+" or "Me too" posts, but the 80 that
remains are not bad at all. Here, I would like Gnus to
only show the good posts. That is probably a
non-trivial task to do, and it wouldn't be perfect,
but I think it could be pretty good given some time
for adjustments. There is a lot of material to train
on so it could be a cool project in many ways :)

-- 
underground experts united

Re: adapt in SCORE file does not work logicaly

[Stefan Huchler]

Emanuel Berg <embe8573@student.uu.se> writes:

> I.e, say some guy posts 100 posts, 10 are flame war
> posts, 10 are "1+" or "Me too" posts, but the 80 that
> remains are not bad at all. Here, I would like Gnus to
> only show the good posts. That is probably a
> non-trivial task to do, and it wouldn't be perfect,
> but I think it could be pretty good given some time
> for adjustments. There is a lot of material to train
> on so it could be a cool project in many ways :)

Hmm you can also Lower by Body substring or regex, so you should be able
to lower the scores by what u want.

L b s p (Lower body substring permanent) or instead of s I think r for
regex.

I don't really know what u miss you should be able to make very good filters.

adapt in SCORE file does not work logicaly

[Stefan Huchler]

the adapt variable in score files which u can set to ignore/nil/t works
as far as I understand it as a group specific version of
gnus-use-adaptive-scoring which is the global version of it, right?

So u can activate gnus-use-adaptive-scoring globaly and with
adapt=ignore dont use adaptive scoring for this group, and the other way
around.

So far that makes much sense to me, but score does not allow the value
word or (word).

So 

1. question: how can I do activate word based adaptive scoring for one
group?

2. question: it would be logical to either allow the value (word)
there. Do I have to quote it somehow or something?

If I press 2 in gnus-score-customize on the score value menu which
stands for lisp expression, it just ignores it, it does not ask me for a
lisp expression.

thx for an answer!