Fig. 6From: Sequence-to-sequence modeling for graph representation learningTwo-level attention mechanismBack to article page