Relation-aware graph attention network for intelligent retrieval over compact domain knowledge graphs
How well can graph-neural retrieval work when the knowledge graph has only a few hundred entities? Specialised engineering domains—aquaculture processing, pharmaceutical formulation, precision manufacturing—produce compact yet semantically dense graphs that starve collaborative filtering and defeat keyword rankers blin...