Abstract

The purpose of this study is to explore the feasibility of constructing RDF-based knowledge graphs by applying generative AI to oral life history data. For this purpose, the oral life histories of seven Korean nurses dispatched to Germany were analyzed and processed through generative AI to automatically produce RDF triples. These triples were subsequently integrated to form a unified knowledge graph. The findings are as follows. First, generative AI was able to produce RDF triples from unstructured oral life history data, thereby demonstrating the possibility of knowledge structuring for such narrative sources. Second, limitations were observed in the AI’s ability to consistently generate predicates with identical or semantically similar meanings, indicating the necessity of establishing a consolidated predicate framework for effective RDF construction. Third, by integrating the RDF triples into a unified knowledge graph, it was possible to interlink individual oral life histories and visualize them for multidimensional exploration. However, challenges were observed due to the colloquial expressions, contextual-dependent nuances, and dialectal features inherent in oral history transcripts, which hindered accurate entity recognition and led to unnecessary proliferation of objects. Despite these limitations, this study holds significance as an exploratory attempt to connect oral life history research with knowledge graph construction. Furthermore, it demonstrates the potential to simplify the otherwise complex process of building knowledge graphs through generative AI, thereby suggesting broader applicability of such methods as a data resource in fields such as diaspora studies and migration history.