"Text Mining Phrases"

dointime

Hi,
I recently created a simple dictionary-based process for locating and counting specific words (i.e. for sentiment the label was "positive" and the words were "good", "awesome" etc.) based on a text file with each line including a single word from the dictionary and applying the dictionary to a set of documents.
I would like to replicate the process but count specific multiple word phrases (i.e. "very good", "better than the best", etc.). I assume this involves a different tokenization and specifying the n-grams, but I cannot figure out the correct process.
Thanks for any assistance.