This paper addresses graph anomaly detection on text-attributed graphs by formalizing it as a node-to-neighborhood semantic consistency problem, where anomalies stem from mismatches between textual semantics and topological relationships. The authors propose N2NSC, a framework that uses two complementary fusion paths to align graph topology with textual semantics, enabling large language models to leverage both structural and textual neighborhood information.
- Formalizes TAG anomaly detection as a node-to-neighborhood semantic consistency problem involving textual mismatch or topological deviation.
- Introduces N2NSC, a framework utilizing two synergistic fusion paths to capture correspondence between graph topology and textual semantics.
- Enables large language models to fully leverage both textual and structural neighborhood information for improved anomaly detection.
- Demonstrates consistent outperformance of current state-of-the-art methods across eight datasets.
The proposed approach helps users identify nodes with inconsistent semantics relative to their neighborhoods, addressing limitations in existing GNN-based and LLM-integrated methods that overlook the correspondence between text and topology.