feature request: differentiate between read and automatically marked read

feature request: differentiate between read and automatically marked read

[Eric S Fraga]

Hello all,

a simple feature request follows.  

I use adaptive scoring and rely on it.  It would be nice to be able to
distinguish between articles that I have read and those that have
automatically been marked as read because of their low score.  At
present, any article with a negative score (but above the automatically
read threshold) that has been read and articles marked as read
automatically all use the 'gnus-summary-low-read' face.  It would be
great to have a further level face, maybe 'gnus-summary-marked-read',
one for those beyond the threshold.

I must admit to not having looked at the code to see how easy this would
be to do.  I will try to do so but thought I'd bring up the suggestion
first... ;-)

Thank you,
eric

-- 
Eric S Fraga with org 9.5.2 in Emacs 29.0.50 on Debian 11.2

Re: feature request: differentiate between read and automatically marked read

[Eric S Fraga]

Hey hum; I should have known better: gnus-summary-highlight which can be
customised.  Ignore the noise! <blush>

-- 
Eric S Fraga with org 9.5.2 in Emacs 29.0.50 on Debian 11.2

Re: feature request: differentiate between read and automatically marked read

[dick]

ESF> I use adaptive scoring and rely on it.

You'd be the first person or bot I've heard say this.

From Gnus Manual Section 8.6 "Initial impressions seem to indicate that
it's [i.e., adaptive scoring] totally useless as it stands."

Like most Gnus things, scoring is a bear to configure, but worse, its
default bag-of-words approach would get it wrong most of the time.

Re: feature request: differentiate between read and automatically marked read

[Tassilo Horn]

dick <dick.r.chiang@gmail.com> writes:

> ESF> I use adaptive scoring and rely on it.
>
> You'd be the first person or bot I've heard say this.
>
> From Gnus Manual Section 8.6 "Initial impressions seem to indicate
> that it's [i.e., adaptive scoring] totally useless as it stands."
>
> Like most Gnus things, scoring is a bear to configure, but worse, its
> default bag-of-words approach would get it wrong most of the time.

Adaptive scoring is not a bag-of-words approach.  Usually, you'd use it
to up-score threads you have actively read (automatically, just by the
fact that you read messages in the thread) and down-score threads you
caught up without reading (again, automatically by the vitue of catching
up the summary without reading some messages).  Then, when sorting
threads according to score, you get threads of interest on top and the
others below.

Like for Eric, that has served me well for years.

Bye,
Tassilo

Re: feature request: differentiate between read and automatically marked read

[dick]

> Adaptive scoring is not a bag-of-words approach.

It is, you just don't know it yet.  The default is bag-of-lines, but
setting `gnus-use-adaptive-scoring` to `(word)` is bag-of-words.  In
either case, it's semantic-free tallying.

Far be it for me to instruct others on what works for them.  Just
noting my surprise to the community.

I know of no companies, Google-scale or otherwise, who have stood behind
an algo that opines about the read-worthiness of one's mail unless the
mail is 99.98% probable spam.

Re: feature request: differentiate between read and automatically marked read

[Eric S Fraga]

On Thursday,  3 Feb 2022 at 08:04, dick wrote:
> In either case, it's semantic-free tallying.

It is.  Yes, semantic based logic could arguably be "better" but sorting
without semantics can be useful.  Would I make any claims about it
always working?  No, but what matters (to me) is that it works in
practice for my use case.  YMMV, of course.

> I know of no companies, Google-scale or otherwise, who have stood behind
> an algo that opines about the read-worthiness of one's mail unless the
> mail is 99.98% probable spam.

So what?  Absence of evidence is not evidence of absence... ;-)

-- 
Eric S Fraga with org 9.5.2 in Emacs 29.0.50 on Debian 11.2

Re: feature request: differentiate between read and automatically marked read

[Eric S Fraga]

On Thursday,  3 Feb 2022 at 07:26, dick wrote:
> You'd be the first person or bot I've heard say this.

I've never been called a bot before.  :-)

> From Gnus Manual Section 8.6 "Initial impressions seem to indicate that
> it's [i.e., adaptive scoring] totally useless as it stands."
>
> Like most Gnus things, scoring is a bear to configure, but worse, its
> default bag-of-words approach would get it wrong most of the time.

Well, it works for me.  Yes, I spent some time getting the weights right
for the various actions used for scoring but, at the very least, it does
sort my emails in order of decreasing relevance/importance most of the
time.  And it does a good job of automatically marking as read emails
that I am seldom going to read.

Also works well for some of the newsgroups (e.g. emacs.help and
emacs.devel) and some rss feeds (arxiv).

For the record, my settings are:

#+begin_src emacs-lisp
  (setq gnus-default-adaptive-score-alist
        '(;(gnus-kill-file-mark)
          ;;(gnus-unread-mark)
          ;; (gnus-read-mark (from 5) (subject 30))
          (gnus-replied-mark (from 50) (subject 10))
          (gnus-read-mark (from 30) (subject 10))
          (gnus-cached-mark (from 30) (subject 10))
          (gnus-forwarded-mark (from 10) (subject 5))
          (gnus-saved-mark (from 10) (subject 5))
          (gnus-expirable-mark (from 0) (subject 0))
          (gnus-catchup-mark (from -5) (subject -30))
          (gnus-killed-mark (subject -100))
          (gnus-del-mark (from -10) (subject -50)))
        gnus-default-adaptive-word-score-alist '((42 . 3) ;cached
                                                 (65 . 2) ;replied
                                                 (70 . 1) ;forwarded
                                                 (82 . 1) ;read
                                                 (67 . -1) ;catchup
                                                 (69 . 0) ;expired
                                                 (75 . -3) ;killed
                                                 (114 . -3)) ;deleted
        )
#+end_src

and I've been using these now for a few years.

I should add that I do have some fixed scores as well so, for instance,
emails to me from the PhD students I supervise always score high no
matter what the subject is.

-- 
Eric S Fraga with org 9.5.2 in Emacs 29.0.50 on Debian 11.2

Re: feature request: differentiate between read and automatically marked read

[Uwe Brauer]

[-- Attachment #1: Type: text/plain, Size: 3135 bytes --]

>>> "ESF" == Eric S Fraga <e.fraga@ucl.ac.uk> writes:

> On Thursday,  3 Feb 2022 at 07:26, dick wrote:
>> You'd be the first person or bot I've heard say this.

> I've never been called a bot before.  :-)

>> From Gnus Manual Section 8.6 "Initial impressions seem to indicate that
>> it's [i.e., adaptive scoring] totally useless as it stands."
>> 
>> Like most Gnus things, scoring is a bear to configure, but worse, its
>> default bag-of-words approach would get it wrong most of the time.

> Well, it works for me.  Yes, I spent some time getting the weights right
> for the various actions used for scoring but, at the very least, it does
> sort my emails in order of decreasing relevance/importance most of the
> time.  And it does a good job of automatically marking as read emails
> that I am seldom going to read.

> Also works well for some of the newsgroups (e.g. emacs.help and
> emacs.devel) and some rss feeds (arxiv).

> For the record, my settings are:

> #+begin_src emacs-lisp
>   (setq gnus-default-adaptive-score-alist
>         '(;(gnus-kill-file-mark)
>           ;;(gnus-unread-mark)
>           ;; (gnus-read-mark (from 5) (subject 30))
>           (gnus-replied-mark (from 50) (subject 10))
>           (gnus-read-mark (from 30) (subject 10))
>           (gnus-cached-mark (from 30) (subject 10))
>           (gnus-forwarded-mark (from 10) (subject 5))
>           (gnus-saved-mark (from 10) (subject 5))
>           (gnus-expirable-mark (from 0) (subject 0))
>           (gnus-catchup-mark (from -5) (subject -30))
>           (gnus-killed-mark (subject -100))
>           (gnus-del-mark (from -10) (subject -50)))
>         gnus-default-adaptive-word-score-alist '((42 . 3) ;cached
>                                                  (65 . 2) ;replied
>                                                  (70 . 1) ;forwarded
>                                                  (82 . 1) ;read
>                                                  (67 . -1) ;catchup
>                                                  (69 . 0) ;expired
>                                                  (75 . -3) ;killed
>                                                  (114 . -3)) ;deleted
>         )
> #+end_src

Thanks for sharing, this is an issue that is in my heads for years but I
never took the time to look into it further. However more and more I
feel I need it. Right now I work with registries and mark threads as
IMPORTANT, TODO, DONE etc, but it is not that useful as I expected,
mainly because you cannot restrict the summary buffer to registries 
(a bug we discovered some weeks ago). So may this is a way to go


> and I've been using these now for a few years.

> I should add that I do have some fixed scores as well so, for instance,
> emails to me from the PhD students I supervise always score high no
> matter what the subject is.

That is to say you score according to their email address? I am
extremely interested in such a configuration, since I right now need it.


Would you mind sharing (off list if necessary)

Thanks

Uwe 
 

[-- Attachment #2: smime.p7s --]
[-- Type: application/pkcs7-signature, Size: 5673 bytes --]

Re: feature request: differentiate between read and automatically marked read

[Eric S Fraga]

On Saturday,  5 Feb 2022 at 09:51, Uwe Brauer wrote:
> That is to say you score according to their email address? I am
> extremely interested in such a configuration, since I right now need it.
>
> Would you mind sharing (off list if necessary)

This is actually the one easy bit about scoring.  In the summary buffer,
if there is an email from somebody that you want to have scored high,
type

C-u NNN I a r p REGEXP-FOR-EMAIL-ADDRESS RET

where NNN is the adjustment to the score you want to assign to the
person's email, "I" means "increase score", "a" means "address", "r"
means regular expression for matching (you can use exact matching if you
wish), "p" for permanent storing the score, and lastly the regular
expression for matching emails from that person which will be pre-filled
with the address of the current email in the summary buffer.

For my PhD students, I use 1000 for the adjustment to the score.

If you use L instead of I, you lower the score for those you don't want
to see (or you deem as less important).

-- 
Eric S Fraga with org 9.5.2 in Emacs 29.0.50 on Debian 11.2

Re: feature request: differentiate between read and automatically marked read

[Uwe Brauer]

[-- Attachment #1: Type: text/plain, Size: 1572 bytes --]

>>> "ESF" == Eric S Fraga <e.fraga@ucl.ac.uk> writes:

> On Saturday,  5 Feb 2022 at 09:51, Uwe Brauer wrote:
>> That is to say you score according to their email address? I am
>> extremely interested in such a configuration, since I right now need it.
>> 
>> Would you mind sharing (off list if necessary)

> This is actually the one easy bit about scoring.  In the summary buffer,
> if there is an email from somebody that you want to have scored high,
> type

> C-u NNN I a r p REGEXP-FOR-EMAIL-ADDRESS RET

> where NNN is the adjustment to the score you want to assign to the
> person's email, "I" means "increase score", "a" means "address", "r"
> means regular expression for matching (you can use exact matching if you
> wish), "p" for permanent storing the score, and lastly the regular
> expression for matching emails from that person which will be pre-filled
> with the address of the current email in the summary buffer.

> For my PhD students, I use 1000 for the adjustment to the score.

> If you use L instead of I, you lower the score for those you don't want
> to see (or you deem as less important).

Thanks very much!!

Fortunately the email is offered when running this, oops cumbersome
command. I tried it out with our email of course. 🙃

Your messages are then in bold face (did you fiddle the setting of the
display variable you mentioned earlier?)

BTW I wish that scoring would it a bit easier to set up, what you sent
me I shall copy in one of my important registers (I use copy-to-register
a lot)

Uwe 

[-- Attachment #2: smime.p7s --]
[-- Type: application/pkcs7-signature, Size: 5673 bytes --]

Re: feature request: differentiate between read and automatically marked read

[Eric S Fraga]

On Saturday,  5 Feb 2022 at 11:36, Uwe Brauer wrote:
> Your messages are then in bold face (did you fiddle the setting of the
> display variable you mentioned earlier?)

Gnus has faces for the various levels of scores: high, normal, low; and
categories within these scores: ancient, read, ticked, and unread.  And
then settings for what is a high or low scoring article:
gnus-summary-default-low-score, etc.

And, yes: I had to play with gnus-summary-highlight and added the
condition:

          ((< score gnus-summary-mark-below)
           . gnus-summary-cancelled)

near the end of the list of conditions to use the gnus-summary-cancelled
face.

> BTW I wish that scoring would it a bit easier to set up, what you sent

Well, I guess I find it rather easy as it's either I for increase or L
for lower and then just follow the prompts with "?" available at each
prompt to remind you what the options are.  There are also the V keymap
bindings in summary mode for seeing the effect of scores.

Adaptive scoring is trickier, mind you.

-- 
Eric S Fraga with org 9.5.2 in Emacs 29.0.50 on Debian 11.2