Re: [Classifier4j-devel] Bayesian Case Study

[moedusa <mo...@in...>]

Matt Collier wrote:

> Another little snowball stemming test.  I suppose consistency is the key to 
> the stemming process whatever the outcome.

I am afraid that any text should be first
  a) tokenised (strip markup, if exists or any other symbols and get raw 
'text' out)
  b) cleaned from stop words
and only after that stemmed...