How to classify text with scikit learn?

I have extracted some bigram features(Adjective/Adverb, Noun/Adjective, Adjective/Verb, Noun/Verb) from a POS tagged corpus with the help of it´s POS tags. The bigrams look as follows: [('word', 'word'), ('word', 'word'), ('word', 'word')].

Given this i would like to know how to use the previous bigrams in order to make some supervised classification with SVM algorithm, as i know mostly any implementation of SVM only accepts as parameters numbers. How can i create a feature vectore and make some classification (positive, negative, neutral). Could anybody provide me some example?. I read the documentation of scikit and the example of text classifcation, but i dont get how is he building the feature veacture that feeds the algorithm.

Thanks