h a l f b a k e r yA hive of inactivity
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
Please log in.
Before you can vote, you need to register.
Please log in or create an account.
|
...and yet I can't find it.
In the three email clients I use I can tag
an
item as spam. This is not only used to
send an email into oblivion (after a short
time in a spam folder), but also (in
theory)
to teach the spam filter.
I can take a message that has been
tagged
as spam,
either by my own error or as a
false positive by an agressive spam filter,
and untag it. This removed it from the pit
of oblivion and (presumably) also teaches
the spam filter.
What I can't do is tag an item in the spam
folder and *confirm* that it is spam. It's
something that I could easily do when I
feel like it and I'm weeding out the few
false positives that get through. The data
could
be used to further teach the spam filter.
So that's it. The Spam Confirm Button: a
checkbox and button that you use to
confirm to your spam filter that a
message
it has already filtered as spam *is* in fact
spam. Not something you have to do to
make the system work, but something
that
could help it to work better.
[link]
|
|
I think that I know what you mean, [st3f]. I would have thought that were you to have cleared all items from your spam/junk folder then they would automatically be tagged and deleted immediately. I still, however, have to remove them from Gmail. Hotmail gets rid of them after a few days 'grace'. Are you suggesting getting rid of the grace time? |
|
|
I don't think so. I'm suggesting that during
the grace
time you can, if you wish, confirm that the
item is indeed spam and, in doing so,
improve the accuracy of the spam filtering
algorithm. |
|
|
this crosses my mind every time I look in the spam folder (sometimes finding something that damn well shouldn't be in there). |
|
|
Am I spam enough for you baby? |
|
|
I bet you say that to all the spam. |
|
|
I just don't get why this is usefull, once spam always spam? |
|
|
I got wads of spam fo spam tog I. |
|
|
I'll have some spam without so much rat in it. |
|
|
I Imagine that most spam sorting
algorithms work by scoring various
aspects of a piece of email. Everything
over a certain score is called spam and
is treated appropriately. Everything
under that score is called mail and
appears in your inbox. |
|
|
If the algorithm is capable of learning
(and the filter would have to be for this
idea to be of use) then it will only learn
when directed by human interaction.
For the spam filters I have used, this
interaction only occurs when you spot
spam in your inbox and tell the filter
about it, or when you spot incorrectly
filtered mail in your spam folder. |
|
|
I want the user to have the option (but
not be required to) interact with the
spam filter by confirming that those
items in the spam folder are in fact
spam. |
|
|
This confirmation would not change the
status of the message. It's already
tagged as spam and will be deleted. The
confirmation simply gives the algorithm
another chance to learn and become
more effective in the future. |
|
|
Not having played with any big
server-based spam filters, it is entirely
possible that this is already baked. If
someone has experience of these and
can point me in the right direction,
please do. |
|
|
But seriously, does it not already know what you are proposing to teach it? |
|
|
It thinks it does. It's sometimes wrong. |
|
|
I can tell it when it's wrong -- I've got a
button that does that. I want a way to
tell it that it's right so that the pathways
that made that decision get
strengthened and that similar messages
will be filtered. |
|
|
Now, if I check every message and
never make a mistake pointing out the
mistakes and reinforcing the correct
decisions are entirely equivalent. I don't
check every message and I *do* make
mistakes, so I can use positive
reinforcement to improve the algorithm. |
|
|
Well, I guess it works that way for kids, but for computerprograms? I doubt it. Sorry, you get a fishbone. |
|
|
A few of the commercial spam filters work in exactly this way, using collective Yes/Nos to improve filtering. |
|
|
It sounds like you need a spam rating. Someone was hinting at that earlier, but it's gone now. |
|
|
You start by assigning each piece of e-mail a spamminess percentage. Letters from your current sweetie are 0% spam, letters from an illiterate religious Nigerian Viagra lottery mailbot are 100% spam. Poorly written letters from your sweetie saying she has got religion and moved to Nigeria, advising you to try Viagra the next time you get lucky, or maybe a vibrator, are assigned as you please. |
|
|
The computer learns your assignment criteria, and sorts mail accordingly. You set the levels that you want used in sorting mail into three categories: Mail, Maybe, and Damnable Spam. |
|
|
You read the mail and auto-delete the spam. When you check the Maybe folder, you teach the computer more of your criteria. |
|
|
1. Due to mix up of some numbers and names, we ask that you ... information confidential |
|
|
2. tometa software: enterprise application called
MailCatraz that we need beta testers on. |
|
|
3. we are looking ... you serve as an in-country distributor |
|
|
4. you ... for the company and the balance will be paid into an account we will ... you. |
|
|
So far today. Yes, the need for better spam filtering will rise as the numbers of buzzwords in spam declines. |
|
|
5. Due to ... affairs I can't visit ... that is
why I have ... you. (probably somebody's little girl) |
|
|
well, I would ease on the interaction by only stating that what ever
was left in the spam folder would be used as a *silent* confirmation.
That is, if he was taught right (and who would question my methods of
teaching?) then most of the time I wouldn't care to reassure it. He
has to have some confidence... |
|
|
I like this idea, because if my sapm filter would actually learn and stop sending me the same crap all the time, then I would have no need for this. But as it is....my spam filter sucks. So I like this idea, then I could confirm that it is really spam and never have to see it again. |
|
|
Gets my vote, too. However, the filter still lets way too much through, and I constantly run the risk of missing genuine emails, even when I pore through the Spam folder to find stuff that has been misclassified. |
|
|
Perhaps you could underline especially spammy paragraphs, words, and phrases for it to spot. Often, when I do start getting more spam than usual, that spam follows a new pattern. |
|
|
Good for getting out aggressions too. Who needs a punching bag? [+] |
|
|
I think most online systems use the less interactive method of looking at *how many* people tag the same message spam [I have a nagging suspicion that this is the main driving force behind Google's almost perfect spam filter] |
|
| |