You expose a series of paper negatives until one is properly exposed and one that will scan properly to form a good reversal image. At that point, you have the ISO. Then you compare that ISO (In-camera) to the paper ISO and begin constructing a conversion table.
OTOH, you have to remember that this is not a true emulsion ISO, as the paper picks up speed from back reflections.
An alternative method, which is much harder, is to do exposures on a spectrosensitometer and then integrate the area under the curve.
I have used both methods. I find therefore that Ilford MGIV paper is close to a true ISO of 25, all things considered such as the high average contrast and the back reflection. I have posted some of that here before.