Experimental study demonstrates context-aware text answering using local retrieval-augmented generation for museum exhibitions, indicating a framework to enhance visitor engagement.
This paper presents a development in constructing the Avatar system based on artificial intelligence (AI) technology to answer the textual questions in its initial phase. The proposed solution is developed using the Ollama tool and implements a Retrieval-Augmented Generation (RAG) architecture. The advantage of this system is that it works locally and ensures responses are strictly based on a given document, which prevents imaginary answers. Use case is given based on the exhibition catalogue dedicated to Serbian mathematician Mihailo Petrović Alas as the knowledge source. The four systematic experiments are conducted with varying chunking parameters, retrieval methods, and system configurations. These experiments are evaluated using 25 test questions across different system setups. Based on the results, research establishes a foundation for developing reliable, AI text answering context-aware systems as the initial phase of creating AI avatars that can enhance visitor engagement.
No takes yet. Share an insight, caveat, or question.
Tatić et al. (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: