What to use a node feature on the hinSAGE algorithm?

Question

I have a graph with this info:

StellarDiGraph: Directed multigraph
Nodes: 1417, Edges: 1988

Node types:
ParagraphA: [429]
Features: float32 vector, length 1417
Edge types: A-contains->D, A-contains->F, A-contains->E
PrecedentB: [332]
Features: float32 vector, length 1417
Edge types: none
GesetzbuchC: [225]
Features: float32 vector, length 1417
Edge types: C-contains->D, C-contains->A
AbsatzD: [223]
Features: float32 vector, length 1417
Edge types: D-contains->F, D-contains->E
Judgment: [100]
Features: float32 vector, length 1417
Edge types: Judgment-refers to->D, Judgment-refers to->C, Judgment-refers to->F, Judgment-refers to->A, Judgment-refers to->A, ... (1 more)
SatzE: [55]
Features: float32 vector, length 1417
Edge types: E-contains->F
NummerF: [53]
Features: float32 vector, length 1417
Edge types: none

Edge types:
Judgment-refers to->A: [472]
Weights: all 1 (default)
Features: none
C-contains->A: [429]
Weights: all 1 (default)
Features: none
Judgment-refers to->B: [350]
Weights: all 1 (default)
Features: none
Judgment-refers to->D: [265]
Weights: all 1 (default)
Features: none
A-contains->D: [221]
Weights: all 1 (default)
Features: none
Judgment-refers to->F: [75]
Weights: all 1 (default)
Features: none
Judgment-refers to->E: [65]
Weights: all 1 (default)
Features: none
D-contains->E: [40]
Weights: all 1 (default)
Features: none
D-contains->F: [28]
Weights: all 1 (default)
Features: none
A-contains->E: [15]
Weights: all 1 (default)
Features: none
A-contains->F: [15]
Weights: all 1 (default)
Features: none
E-contains->F: [10]
Weights: all 1 (default)
Features: none
C-contains->D: [2]
Weights: all 1 (default)
Features: none
Judgment-refers to->C: [1]
Weights: all 1 (default)
Features: none

I want to use hinSAGE to get a node embedding of only the "Judgment"-type nodes and use the other types as information as it is described in the docs of hinSAGE. One problem that I have with my graph is, that I dont have any node features. However I calculate a vector representation of the judgment text bevorhead. Is it a good Idea to use this vector representation as a node feature for the "Judgment"-nodes?