Yeah, that's why I can never understand any of those silly letters you send me

[venta]

For those of you who are still occasionally amused by auto-thesaursusised spam. I know I am :)

These lozenges are merely equal standard lozenges but they are specially developed to be spoiled and soluble below the lingua. The pills is took up at the rima oris and goes into the bloodstream in real time alternatively of moving through with the breadbasket. This effects in a quicker much more mighty effect which run up to 44 hours!

Comments

[mejoff.livejournal.com]

auto-thesaursusised? it looks more badly translated from the [insert oriental language here]. Some googlesearching should reveal vast amounts of such, mainly instruction manuals for electronics. I love them, they're my favourite!:)

[mejoff.livejournal.com]

www.engrish.com

[venta.livejournal.com]

From less convoluted examples, I'd come up with the idea that it might be auto- (or just badly-)thesaurusised to avoid simplistic spam filters.

If I get mail, of only a few sentences, which contains the words "tablet" and "hard" and "penis", then it's a reasonable bet it's spam. If it contains "pill" and "rock" and "member", then the filter doesn't trigget. So, I then add those words to my filter... so the spammer starts using different synonyms. Which get less and less synonymous as time goes on.

I haven't done any looking into this, it was just my theory. Admittedly, though, at least some of the above example shows signs of bad auto-translation as well.

[lanfykins.livejournal.com]

I am reminded of the old BabelFish game.

Comes, friendly bomb, and autumn now is not the grass grazes the cow group in its not suitable manner there mire, the death

[venta.livejournal.com]

Aye, I was just remembering that. I may knock up a quiz based on that principle when I've the time.

[wimble.livejournal.com]

I'm using (well, I was: my connection to home has fallen over!) spamassasin, which has a Bayesian logic filter in it. Which does essentially the same job as you're describing, in adding specific words to the spam filter.

But it has the advantage of adding all the words in the spam message, giving them automagically adjusted weightings. So even lingua goes in.

When you add this to the normal spam detection rules (duff dates, silly html colours, "Dear friend", etc), and finally, the network spam centres (DCC, Pyzor, which hold checksums of spam mails, in much the same way that freedb does for CDs), it rapidly adapts to changing content.

So far, this year, 44 spams have got through the filter. I don't know how many it's blocked.

[onebyone.livejournal.com]

Do you know how many false positives there have been?

[wimble.livejournal.com]

Zero :)

My alledged spam goes into a "caughtspam" folder, which I periodically browse through and then delete.

All the messages scoring above 5 are classed as spam, and those scoring above 10 are automatically added to the bayesian database. When I flush the file, I tell it to learn all the rest (ie. those between 5 and 10), which is usually about half of the content.

Ah... It also auto-whitelists addresses, so recognised non-spammers are given a certain amount more leeway in their content. And that really does improve things (given that I get quite a lot from LJ Notify!)

[venta.livejournal.com]

I don't know how my Spam filtering works - it's done by my mail provider. I don't bin spam, though, just mark it as (a) I'm worried about false positives and (b) I quite like reading the bilge from time to time. I think I've had three bits of spam unmarked since I got the blacktreacle address (nearly a year).

My notification of my vodafone bill used to get marked as spam, but after a couple of cases of me saying "no, no, I want this" it seems to have got itself sorted out.

[wimble.livejournal.com]

I've never liked the idea of not knowing, and the filtering being done by the provider: they might (potentially, for the paranoid amongst you) be deleting "spam" before it ever gets to you. And if they've got a false positive, you'd quite likely never know.

I get quite a lot to
a) my ntl address, since it's an obvious domain, and I've got a very short username.
b) directly to my mailserver at home, since it'll simply accept email.

[maviscruet.livejournal.com]

Good lord, your title, I recognise the song.

It is a billy brag song right?

Or am i mad?

[venta.livejournal.com]

One notional kudo to that man. It is indeed Billy Bragg, from Walk Away, Renée.

[zotz]

The most illegible bachelor in town . . .