Let $G$ be a finitely generated group. Let $S(G)$ be the quotient of $\text{Hom}(G,\mathbb{R}) \setminus \{0\}$ by the equivalence relation that identifies two homomorphisms if they differ by scaling by a positive constant. For a nonzero $\phi \in \text{Hom}(G,\mathbb{R})$, write $[\phi] \in S(G)$ for the associated equivalence class.
The Bieri-Neumann-Strebel (BNS) invariant of $G$ is a certain subset of $S(G)$. The definition in the original paper is hard to parse, but in several places I have seen the following definition of it:
Fix a generating set $S$ of $G$. Let $\text{Cay}(G,S)$ be the Cayley graph of $G$ with respect to $S$. Consider a nonzero $\phi \in \text{Hom}(G,\mathbb{R})$. Define $X_{\phi} = \{\text{$g \in G$ $|$ $\phi(g) \geq 0$}\}$. Then $[\phi] \in S(G)$ is in the BNS invariant if and only if the full subgraph of $\text{Cay}(G,S)$ spanned by $X_{\phi}$ is connected.
I have seen it asserted without proof or reference that the above property does not depend on the choice of generating set $S$.
Question 1: Can someone give me either a proof or reference for this?
Question 2: Can someone explain how to relate this to the definition in the original paper defining the BNS invariant? A reference here would also be fine.