This has been the most hectic travel for conferences. I've done two conferences in a week before, Chicago and Taiwan, this time New York and Berlin. But the pressure this time is much higher - doing the tutorial and presenting at ACL.
The tutorial NLP in MIR: I think one of the questions that gets asked a lot is how NLP is integrated to MIR. In other words, what makes it special of doing NLP in MIR, vs. using some data from other domain? Certainly that should go beyond asking word2vec to answer "Dylan is to guitar as Mozart is to ___" and come up with the answer 'piano'. To me, there are different levels of integration. For some tasks, text analysis is just text analysis but they may provide additional information for content based MIR. Yet in others, maybe it is a way of thinking, the bread and butter of doing NLP and how that could inspire people to do new work in the MIR domain with non-textual data. And yet another dimension is some music domain specific NLP techniques to serve the integration, such as relation extraction for music domain knowledge bases. For me, there is endless ways to consider this integration, and we've touched on many different ones (in my topic modeling case, I emphasize the way of thinking in language modeling, for example, really working through a way of thinking about the generative story; and in Sergio and Luis's, they do more on the music side). All in all I think it is a good tutorial with many different facets.