A Context-Aware Cultural Heritage Guide Powered by LLMs

2026-08-01

AI Summary

This study addresses the limitations of insufficient depth and personalization in cultural heritage interpretation by proposing a loosely coupled, large language model (LLM)-agnostic, context-aware architecture that dynamically integrates external information relevant to either the user or the exhibited artifact. By decoupling the LLM component, the approach enables flexible model substitution while ensuring system stability, and it has been implemented within the Triangolazioni tour system to support contextualized content retrieval and generation. Experimental results demonstrate that this architecture significantly enhances the relevance and richness of interpretive content, offering a scalable and highly adaptable paradigm for augmenting digital experiences in cultural heritage contexts.

Abstract

We present an extension of Triangolazioni (a Cultural Heritage webapp) to enrich curated content with context-dependent, external information provided by Large Language Models (LLMs) within a loosely-coupled architecture agnostic to the LLM. The system supports context-dependent information search and presentation within an architecture agnostic to the exploited LLM.

Problem

• Cultural Heritage
• Context-Aware
• Large Language Models
• Information Enrichment

Innovation

• Context-Aware
• Large Language Models
• Cultural Heritage
• Loosely-Coupled Architecture
• Information Enrichment