Note: I am involved with this project and can answer questions. As might be expected, the article makes things sound further along than they actually are.
Analytics meets the Academy Awards...The "People's Oscar" tool co-developed by the L.A. Times, IBM and the University of Southern California Annenberg Innovation Lab analyzes Twitter opinions about who's going to win Academy Awards. An 'entertaining' example of how semantic analysis is performed on large set of unstructured data/tweets.
We use QM Perception extensively for evaluating learning/capability and deploying surveys. Most valuable to our analysis is the Item Analysis report which has extensive data and built in data for psychometric analysis.
I am not sure where there is direct relevance to Learning Analytics. There is no doubt that mining the public data of social network will be useful for identifying trends, possibly with implications for learning. The big unknown is how reliable is it? Are these networks representative of a population or specific group? I am not convinced of that.