Re: subject - why not all caps?

[ha...@t-online.de]

>> At 12:38 PM 7/9/2005, hamann.w@t-online.de wrote:
>> 
>> 
>> >I just received this spam (some of them really get their stuff translated 
>> >well now) but was
>> >surprised that it did not trigger subject all caps rule
>> 
>> 
>>  From the eval test code for that rule:
>> 
>> 
>> # don't match one word subjects
>> 
>> 
>> Since that subject only has one word, it would have missed.
>> 
>> Besides, SUBJ_ALL_CAPS isn't enough score to be worth worrying over.. it's 
>> less than 1.0.
>> 
>> 
>> 

Hi Matt,

thanks for your reply.
It seems these guys know how SA treats their messages :(

There have been discussions before about "amplifying" rules, but I am not sure whether
it is possible. I would say that both english and german versions of this
"BUSINESS PROPOSAL" share a few characteristics that could be combined like a meta
rule
- subject all caps
- at least one run of all caps words somewhere in the body
- mentioning dollars (is detected)
- mentioning specific places (Lagos and Abijan are common - there are spelling variants on Abidjan)

As for the all caps rule, it is hard to understand why it was written not to fire on a single
excessively long word.

Wolfgang Hamann

Re: subject - why not all caps?

[Loren Wilton <lw...@earthlink.net>]

> As for the all caps rule, it is hard to understand why it was written not
to fire on a single
> excessively long word.

I'll take a guess at that one:

Single word caps subjects are likely to be an acronym or similar, and are
moderately likely, at least in US business mails.

I suspect the 'excessively long' part was never considered when writing the
exception.  However, it is probably just as well that it wasn't, because
compared to English, German (just by the nature of the language) has a great
number of 'excessively long' words.  That sort of rule would probably be
causing FPs all over the place.

My opinion is that it is probably time that we should be putting together
the basis of a German spam rule set.  There may not be a need for a huge
number of rules yet, but it seems clear that German-language spam is
starting to show up.  It will doubtless only get worse with time; after all,
Germans have a considerable amount of money they can spend, and that is what
spammers are after.

For instance that message that someone (you?) posted looked to me like it
was probably a form of a Nigerian spam.  Those can be a bit slippery to
catch, especially when not done in all caps as they used to be.  But there
are a number of stock phrases that show up in them, and you can build a meta
that will catch 3 or 4 of them together, and be almost certain it is a spam.

What someone would need would be 3 or 4 examples of this sort of thing to
start seeing what were the common words and phrases, and then we could start
building simple rules to look for those phrases.  Of course, understanding
the language would make this task considerably easier!  :-)

        Loren