Swoogle is a semantic search engine project by the Computer Science and Electrical Engineering Department at the University of Maryland. They have taken a useful approach by ripping off Google's interface so that new users understand how to use the tool from the very beginning. Unfortunately the major limiting factor of the experience seems to be the results page. Rather than creating human readable snippets of information below each of the links the results simply output a snippet of the RDF code from the returned file. RDF is difficult for computers to understand so asking people to make sense of a brief quote is expecting the possible. As a consequence it is difficult to meaningfully interrogate the results in order to find content that is most relevant to you, in fact in practice the value of the returned results seems almost zero.
I am not sure how semantic search engines (if they ever come into fruition) will look but I do not think they will resemble Google. It seems like the real value of semantic search lies in identifying the user's search context and linking this to their attention stream (something Steve Gillmor talks about a lot). It is way too time consuming to describe many of the semantic nuances you wish to include in your search, and in search engines where you must manually enter this data a 'dumb' Google search is of more immediate value because of its speed. In the time that you describe all your semantic search parameters you could perform a dozen standard searches and probably find twice as many relevant results. If on the other hand the semantic search engine tied into your working context, say for example it new were working on concrete detailing within your CAD file, and it had access to your attention meta-data so that it could determine relevance then the potential for semantic search could be very strong. However whilst these two essential building blocks do not exist semantic search will always be either too time consuming and difficult to use or so simple that it is not as effective as plain old Google approach of using PhD's in tanks to get through walls (from Adam Bosworth's MySQL conference speech).