Have you also checked out Qanda? I think that the reason why most of these projects are dormant/abandoned is that the real difficulty for a question answering system does not really lie in algorithms and coding, but rather in building a knowledge corpus, which involves a lot of painstaking labor.
Also, results will tend to be much better if the system is domain-specific rather than general purpose, because you can tailor the corpus for a particular set of subjects.
