Automated Graph Production of Museum Commentary using Named Entity Recognition and Relation Extraction

Abstract

In this paper, we propose an automated method for creating graphs for museum explanations using natural language processing technology. For this purpose, we collect data on museum artifact explanations from various museums such as the Gwangju Museum and the Cultural Heritage Administration to learn Named Entity Recognition (NER) and relationship extraction models. The KoBERT model is used to learn the NER model, and the rule-based model is used to extract relationships. Based on the entity information and relationship information extracted using the learned model, the information in the museum is mapped to the ontology and organized into a graph. Then, experts and non-experts compare and evaluate the graphs created by the AI from the perspectives of museum visitors and graph creators, respectively. The graph constructed through the proposed method provides information that can be easily understood by users. It helps to organize and maintain the knowledge inside the museum and can be used to form the database of a chatbot or recommendation system.