Will Hobbs

@BenListyg Agreed on points above. A lot of the methods for short text, including mine, boil down to analyzing a more aggregated and dense version of the data. Hand coding / a lot of reading probably helps if you're thinking to bring in priors.

@ any / all NLP twitter people: What's the state-of-the-art for topic modelling with "small" documents (e.g. text messages ~50 words)?