Technical analysis reveals accuracy limitations of virtual assistants in cultural heritage databases, highlighting the need for retrieval-augmented generation and fine-tuning.
After years of work on cultural heritage and scientific digitization, the world accumulated a vast amount of data and information. Although digital data is far more accessible than the physical one, even in an unstructured way, better interaction is needed by the non-professional users. At the same time AI-related systems gain widespread popularity in all kinds of platforms, easing accessibility. Digital assistants and chatbots are implemented in existing applications and databases. Cultural heritage areas are not falling behind. Plenty of examples of using such systems in the field are already available. Implementation of AI in cultural heritage brings unique challenges along with benefits. The accuracy of the models is highly uncertain, caused by the specifics of the cultural heritage data. Several approaches to overcome the challenges such as parameter-efficient fine-tuning and Retrieval-Augmented Generation are examined.
No takes yet. Share an insight, caveat, or question.
M. V. Kraev (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: