Enhancing Access to Movie Cultural Heritage through Narrative-Enriched Item Representations

Movies are an important form of cultural heritage, encoding social values, narrative traditions, and collective memory over time. However, digital platforms that mediate access to movie collections - including recommender systems, digital archives, and streaming services - typically rely on explicit metadata such as genres and plot summaries, which only partially reflect the semantic richness of cinematic content. In this paper, we propose an approach to augment item representations with implicit narrative information extracted from textual descriptions to provide recommender systems with richer item profiles. Building on an LLM-based method for extracting hidden narrative aspects, their interpretations, and their manifestations from item descriptions, we show how these enriched representations provide more expressive and semantically grounded descriptions of movies. We report the extraction performed on the MovieLens 1M dataset. We further discuss how these narrative-enriched representations extend beyond recommendation performance to support deeper cultural exploration, thematic discovery, and interpretation of movie collections, addressing key challenges in personalization for cultural and natural heritage environments.