Automated solution for sifting out most of the spammers?

I had been thinking about an automated method for sifting out most of the spammers.

Here is what I see (not sure if it's prohibitive due to too many API calls somehow):

1) Let the user enter threshold values for the 3 values used below,
as well as possible exclusion keywords (e.g. "liberal", "conservative", "God", etc.)
to check against the bio text.

2) For each new follower, fetch their extended info (1 call, no?), parse the XML,
and then do the following checks on it

foreach ($exclusion_word as $ew) { if (preg_match('!$ew!', $biotext)) $exclusion=1; }

3) if OR $exclusion=1

// people don't follow them back
OR friends / followers > 2 (default, your user can adjust to his/her liking)

// they have few updates, as well as 10 x the followers than updates, -> "spammy"
OR (updates < 50 AND followers / updates > 10)

THEN - Do NOT follow (i.e. "continue"), and possibly put them in a "quarantine" email
for the manual processing you just developed.

ELSE follow

------------------------------

Do you think this could work?

Best - Alex Schleber