I think the fundamental problem with the Semantic Web idea was that it was way too much work, not only to do the actual annotating, but to work out what the system would be for a given domain. Endless, years-long back and forth between experts that usually ramified into an incomprehensibly complex system just to establish the language, let alone to use it. But the real problem is that such systems set in stone the way one particular group at one moment in history see the world. Imagine how 19th C scholars might have set this up for history, art, or (pick your subject!) Any such system embodies a moment in time and has a half-life of at most a generation, a time span roughly equal to how long it takes to work out in the first place. The structure, if not the data proper, will inevitably go out of date faster than it can even be established, and leave an almost un-moveable mass of steadily increasing size for the following generations. It's basically the same thing that happens when you get old--all the millions of things you know become increasingly irrelevant and there's too much to shove aside to make room for new mental structures.