There were some folks looking into building RDFPath, an XPath-like syntax for referring to RDF data, although it hasn't had any updates for over a year. The syntax is fairly general purpose though.
For the RPath language I'm using in the SWC, I probably have different goals, since the expressions are intended to be used in UI elements, like XPath expressions are used in XForms. As such, they don't connect to a giant database of info.
For example, I might use:
director[birthdate/year < 1950]
which in RPath would return all nodes of directors born before 1950, but that wouldn't make sense to use on a client, as a client wouldn't have a database of directors to scan through unless it sent the query off to some remote server.
But, given two RDF files sent to the client each describing a different movie, one could pick out cast members that starred in both using:
children(<movie1id>/cast) & children(<movie2id>/cast)
Thus, there needs to be a balance between making the expressions powerful, yet ensuring that they make sense and performance doesn't suffer as a result.