Abstract: Introduction. This study addresses the representation of ethnic minority cultures in online museum collections, which often reflect diverse viewpoints. We propose a data-driven methodology to construct a large-scale multi-viewpoint knowledge graph, using Jewish cultural heritage as a case study.
Method. We developed an LLM-based pipeline that combines object typing, named entity recognition, relation extraction, enrichment, and clustering.
Results. An analysis of 647,951 records and 178,444 extracted subjects from the collections of Jewish museums across the globe revealed diverse thematic emphases: Israel and the Netherlands prioritised religious themes, while others highlighted everyday life. Surprisingly, only Australia emphasised the Holocaust.
Conclusion(s). The central contribution of this study is the development of a knowledge organisation system capable of tracing major trends and identifying patterns in the polyvocality of perspectives. The methodology provides quantifiable, scalable analysis of multi-viewpoint cultural heritage, extendable to other minorities.
