Physicists computing multiloop Feynman diagrams have introduced various techniques and conjectures that involve the notion of Degree of Transcendentality (DoT). From what I understand one defines
1) $DoT(r)=0$, r rational
2) $DoT(\pi^k)=k$, $k \in {\mathbb N}$,
3) $DoT(\zeta(k))=k$,
4) $DoT( a \cdot b)= DoT(a)+DoT(b)$
One then proves for example that the $\ell$-loop contribution to a certain scaling function in $N=4$ Supersymmetric gauge theory consists of a sum of terms all of which have DoT equal to $2 \ell-2$.
This can't be rigorous mathematically, since it is not even known that $\zeta(2n+1)$ is transcendental, but is there some circle of ideas, or conjecture in mathematics that if true would give a precise definition to DoT?