In principle, it can be done. We know the entropy of all elements in its standard state, so the measurement boils down to measuring entropy differences between the pure elements and the compound or mixture at some given temperature and pressure (or whatever variables define the equilibrium state in question). Now you can set up a reference process in the course of which a given amount of entropy is created. Now you can test out how much of this reference process you need minimally to convert your pure elements into the compound in the given state. If the entropy for this process is negative, you have to reverse your standard process. So in principle, you can test how much of a standard entropy process you need just like you test how much of different weights you need to bring a balance into balance when you determine mass.
While this is theoretically sound, in practice, it is usually completely imperformable, but this should not come as a surprise. It is also completely impracticable to determine almost any absolute temperature using a reversible Carnot engine, nevertheless, it could be done in principle, and this is all what counts.