Synergizing structure and semantics: a knowledge graph-transformer framework for narrator disambiguation in hadith networks

Abstract Historical transmission chains (isnads) are fundamental to verifying authenticity in Hadith literature, yet narrator identity resolution is a persistent challenge due to onomastic ambiguity and complex naming conventions. While traditional methods lack scalability and modern language models overlook crucial network structures, this study bridges the gap by synergizing structural and semantic information. We introduce a novel hybrid framework that integrates a Knowledge Graph (KG) representing the narrator network topology with a Transformer-based model for deep contextual understanding. Our approach first leverages the KG to generate a high-probability set of candidate identities, then employs a hybrid scoring model to evaluate them based on both global network prominence and local semantic compatibility. Evaluated on the AR-Sanad 280K-v2 benchmark, our method establishes a new state-of-the-art, achieving 97.8% accuracy and significantly outperforming existing baselines. This work provides a scalable, high-fidelity solution for narrator disambiguation, advancing computational methods in Hadith studies and historical identity resolution.