Add compression to static corpus data, use chunked computation, hide read corpus ids in corpus analysis