This sounds like a collaborative filtering problem. But rule based system
alone might not be your best choice for such a dynamic environment like
twitter. I would say if u can develop a bag of word approach to write a
classifier and add that to your rule based system then u stand a good
chance, I would assume. I would assume a few hours worth of tweets from
Stream with classification done would serve a good training set for the
algorithm. I do not have any empirical evidence as of now, but that my hunch
about this.


On Sat, Feb 27, 2010 at 6:29 PM, Fabien Penso <fabienpe...@gmail.com> wrote:

> Hi,
> I'm currently using the streaming API for a new service I work on, but
> I see lots of tweets I would consider as SPAM and I'd like to find a
> way to prevent it.
> I have not found anything to filter them, therefor I wrote a little
> blog post about how it could be done. Something to combine RBL and
> Tweets, but I wonder if that makes sense.
> Any feedback welcome.
> http://blog.penso.info/2010/02/28/filtering-spams-on-twitter-twitterbl/
> --
> Fabien Penso
> @fabienpenso

Atul Kulkarni

Reply via email to