Estimating corpus size via queries