To your point about tags, that's why I also mentioned this:

along with associated probabilities that the predicate actually belongs to each of the specified domains.

In other words, the tag originally assigned by the creator of the information can be rated as not applicable. Similarly other people could propose alternative tags that could then be rated.

Of course, none of the above argues against having an option for an automatic tagging system as a supplementary way to check (or even identify) domains as you're suggesting (and its ratings could also be rated by people). So here we could see cases where humans and AI are both checking each other's work.