tm corpus stemming doesn't look right

Question

I have a corpus containing a bunch of tweets, which I have to classify per topic. I\'d like to stem them in order to work better with DocumentTermMatrix and