But with the Open Graph protocol you do get the ability to tie disparate pieces of data together. For instance, let's say that I examine someone's tweet stream for links. If I fetch the OG data for each of those I can tell you whether they tweet most often about people, or movies, or whatever. I can show you a thumbnail display of images representing the things they linked to.

Open Graph is most definitely semantic web. Liking is just one possible application of that semantic data.

Oh yeah - I definitely agree.

I guess my thought was just that the semantic lens is focused in a different direction that I imagined. The mental model I always had was that the semantic info would let users make connections between datasets that model our physical world rather than datasets that monitor users themselves and how they use the web.

This is just a different use case than I imagined.