Since the beginning of this semester at Maryland, I've attended several companies' talks: MS, Amazon... And the Twitter's talk today I think is the most open and technical oriented one. The talk is delivered by the Chief Scientist Twitter: Abdur Chowdhury
The twitter team is working on data mining main trends and hot topics using the huge scale twitter dataset. That's cool indeed. The main problems are entity extraction, local area segmentation, language detection and so on.
I think as twitter is a large scale social network, what we can do should not be limited within recording what already happened, but a reasonable prediction also can be done, such as social riots, or social "earthquake". People do talk before they do something and there must be some rules we can learn from those data to do predictions, I believe. Of course, it may be a very challenging work.
And finally, got a twitter T-shirt~ Hooray~