OK, we're getting somewhere with WhoseKidAreYou. 865 hits off Hacker News. Thanks to comments, it looks like DBpedia already has a ton of data we can use - their data class Person includes a subclass Relatives, and their API uses SPARQL, so a lot of WKAY might just be a frontend to that, probably with a caching element. I envisage caching the most-trafficked classes, and routing everything else to a view that does a lookup against DBpedia dynamically.
This is probably quite an important point about "crowdsourcing" - as the crowd is made up of volunteers, you really, really, can't ask them to double-handle anything, so it's vital to reuse everything.
We'll also need to deal with the case where the Wikipedia record doesn't have any relatives, or when it doesn't exist at all; I'd like to return an embedded editor and get our users to fill in the gaps. We could perhaps pass the result to another user to peer-review before submitting. The client-side/browser stuff needs to wait until the core system is ready.
Various people have asked how they can help - the short way is to join the group. Send me word at my e-mail address or in the comments and I'll send you an invite.