There are countless questions about how an algorithm would work best. I think you took some good first steps in showing how you could retrieve pieces of an entry – the really interesting next step would be to assemble some manual queries for a couple of real-world requests – i.e., pick a few breaking stories, think about who you’d want as a source, and walk through a process to conduct Google Scholar and Google News sources to surface a list of experts, then see how your system did in meeting your needs. Next step is to do the same thing for someone else and react to his/her feedback. All of this will end up shaping the prototype and wireframes, and bringing you closer to something you can propose for funding.

I think this is something that’s a few turns of the crank away from something fundable, but I think you could get closer by tuning the algorithm yourself, manually, before turning it into software. Would be excited to see this go further.