What information can be derived from the newest Social Media crazes?

It is unstructured, streaming, and highly demographic.  Can we effectively mine any value from random thoughts, dribbles, posted pictures, acronyms, and emoticons? 

I published a paper 10 years ago on multi-media indexing.  Think of what could we do with this data from pictures, video, emoticons, and the correlations to text in posts?  Imagine being able to search for pictures where 3 people are wearing bunny ears and saying, "OMG partayyy goin on. LMAO".

Is there master metadata on LOL vs lol, ROFL, TTYL, etc.  Who runs data governance on acronym lingo?

Is this data practical?  Do we really need to know where Miley Cyrus or Ashton Kutcher are eating lunch, or do we get more value out mining the data to see who reads their posts? 

Bottom line:  Seems like something valuable for companies that market to a younger demographic (21 years old and under) based upon what they are subscribed to verses what the posts actually mean.