From fragmented narratives to spatio-narrative networks: a knowledge graph approach to Tianjin Grand Canal heritage

Linear cultural heritage derives meaning from relationships among people, events, places and historical processes, yet these connections are often fragmented across textual sources. Taking 98 heritage sites in the Tianjin section of the Grand Canal as a case, this study constructs a narrative knowledge graph from multi-source texts using LLM-assisted extraction, event hierarchy modeling and graph-database construction, and projects it into a site-level narrative network. Network and spatial analyses identify narrative visibility, thematic clusters, endpoint roles and potential narrative corridors. Results reveal a comprehensive Old City–Sanchahekou core, specialized hydraulic units along peripheral canal sections, locally embedded memory in Yangliuqing, and medium- and weak-strength relations that connect these components. By linking traceable narrative evidence with spatial network analysis, the approach explains how dispersed heritage sites form an integrated relational structure and provides an evidence-based basis for holistic interpretation, thematic presentation, route organization and public communication in linear cultural heritage.