This is not final. I think the content generator will ultimately operate on a preprocessed version of the tweets JSON data from the Twitter archive.
