Describes an approximate document ranking process that uses a compact array of in-memory, low-precision approximations for document length. Combined with another rule for reducing the memory required by partial similarity accumulators, the approximation heuristic allows the ranking of large document collections using less than one byte of memory…