At day 2 of this year's ACM Multimedia, a things are worth to mention here.
First, there was an interesting talk from Yahoo! Research about resolving tag ambiguity where they measure ambiguity of photo tags as a difference between distributions and perform some optimizations. User tests show that 20-26 terms are enough to “compute” ambiguity. They've found different ambiguities such as geographic, temporal, etc.
Second, the brave new topics - controversial by definition! - comprised a presentation about social signal processing (SSP). This was first mentioned in the Signal Processing Magazine 2007. One of the (main) open issues is getting psychology and engineering closer (which is already happening by the way). The presentation was coming from a Network of Excellence founded by EC FP7 with 5 years duration! One of the main goals: create THE Web portal for SSP ;-) Basically, the signals are coming from sensors such as microphones and cameras and the aim is to process/analyze the social behavior of users. Honestly, to me, that’s yet another multimedia analysis approach unless the get more sensors involved. But what knows a stranger...
Finally, the social event provided a decent dinner but ended rather early and quickly. We found asylum in a sports bar near our hotel. Our colleagues from Toulouse got the best paper award and Prof. Steinmetz got awarded with the SIGMM award for outstanding technical contributions to multimedia computing, communications and applications. Congratulations!