Something that caught my eye in the weekly W3C Newsletter was the release of EmotionML 1.0 XML schema (link corrected). The main purpose is to annotate emotional reaction within a recording (video/audio, but conceivably text as well), but the other is to define a framework for emotion recognition on video (Hmmmm).
There are some use cases listed on the site as well as the first draft of the markup, but it looks like a psychology degree would be helpful here. Interestingly, a lot of it has to do with concepts like “arousal”, “friendliness”, “dominance”. At first glance, the values seem a little more related to body language (and inferring emotion from body language).
I can see some very legitimate uses for a markup schema like this, but I also have to confess being a little spooked. How accurate will an “automatic recognition” system be and will it hold up in court? Stay tuned, I guess.