The protocol is hardly the issue with setting up your own movie-search db - it's data.

I guess Freebase might be a way to go, once you've resolved the initial dataset with wikipedia, update incrimentally with just the changes.

Or, you know, you have to license the material from somewhere.