bayesian - Android Spam Detection Application? -

i developing spam detection application android,i using bayesian classification detecting spam messages. want know whether should use training set of 50 ham messages , 50 spam messages or whether should user based content training? effect have on effectiveness of application? know might broad discussion precise answer, not discussion here.

it looks you'll need thousands of training messages.

note spammers have discovered ways past kind of filter, e.g. mispellings "v1agra". iterative refinements classifier might catch current techniques.

bayesian_spam_filtering looks place start, esp. references in-depth articles.