I spent a good amount of time trying to build a contextualization engine for a different reason (auto correct, sick to death of it being so bad when I'm clearly

I spent a good amount of time trying to build a contextualization engine for a different reason (auto correct, sick to death of it being so bad when I'm clearly talking in a very specific context bubble using these same words that everyone else is using over and over), and man can I say that is is extremely difficult to get 5 people in a room to not define 5 different context buckets for a specific piece of text in a given situation. In the end I decided that the real world phenomenon I was trying to model was probably closer to a Google+ multiple identifies than it was to a single channel like Facebook.