Well, this is something I'm interested in.
AnsProlog, which is Prolog-esque syntax implemented via answer-set programming (http://en.wikipedia.org/wiki/Answer_set_programming) is a language that's heavily derived from Prolog, and an active area of research. It's solver-based, however (roughly like a SAT solver), which comes with pros and cons.
I used both Prolog and AnsProlog in a project to model and query videogame mechanics. The idea is to prototype a game's core mechanics in Prolog, and then be able to both play the game and query properties of it. We use Prolog to actually run the game, and mainly ASP to perform queries (because it allows adding various constraints). Sticking to a subset it's possible to run the same code in both, with some mechanical syntax translation.
Possibly a somewhat esoteric example, and definitely researchy, but imo Prolog and/or ASP still have promise in a number of areas. Prolog retains some advantages over the newer ASP in that it doesn't ground out to a propositional solver, which can make it faster in some cases (especially in large domains), and can do imperative-type stuff (there are even GUI bindings), while ASP is purely a formal solver. ASP's advantage is more flexible modeling (e.g. you can have elements of constraint programming), more active research community, and cleaner semantics in some areas, e.g. circular references can be solved.
A few sources:
[1] My collaborator's blog post on that project: http://eis-blog.ucsc.edu/2009/06/you-have-to-mine-the-ore/
[2] A paper on it: http://www.aaai.org/ocs/index.php/AIIDE/AIIDE09/paper/viewFi...
[3] An op-ed I once wrote, "Why did Prolog lose steam?" http://www.kmjn.org/notes/prolog_lost_steam.html