Abstract:

The rapid evolution of generative artificial intelligence (GenAI) is transforming the ways users interact with digital systems, fundamentally reshaping expectations for adaptability, personalization, and engagement. As digital environments continue to expand, so does the demand for intelligent software capable of providing dynamic and context-aware user experiences. This shift presents particular challenges for small development teams, especially those operating within academic institutions, which often lack the extensive resources, infrastructure, and workforce available to leading technology corporations. Despite these constraints, such teams are increasingly expected to deliver innovative solutions that remain competitive and relevant in fast-changing technological landscapes.

In response to these pressures, this paper introduces an AI-based solution developed at the Faculty of Information Technologies (FIT), University Džemal Bijedić of Mostar, in partnership with the Centre for Peace and Multiethnic Cooperation (CZM). The project addresses the broader goal of enhancing the digitalization of cultural heritage through the strategic and responsible implementation of GenAI technologies. The system was tested using partially prepared historical materials provided by CZM, formatted for a TV-style historical calendar and written in the Bosnian language. This dataset served as a realistic and domain-specific foundation for evaluating the system’s capabilities and limitations.

To ensure that the AI operated within appropriate boundaries and remained aligned with the linguistic, cultural, and historical characteristics of the source material, several modules were intentionally designed to function offline. This decision supports both data protection requirements and the need for precise domain-restricted processing. The resulting system reflects a set of functionalities that mirror essential curatorial responsibilities within an e museum environment. Among these modules are automated cross-checking of historical data, extraction of geographical locations, intelligent image selection based on contextual relevance, metadata extraction and harmonization, categorization of materials, evaluation and transformation of data quality, and the generation of questions and answers aligned with predefined learning outcomes.

Furthermore, the system provides support for higher-level curatorial activities, including conceptualization, planning, and selection of exhibition content. By integrating GenAI into these processes, the prototype demonstrates how technologically supported workflows can strengthen cultural heritage preservation, particularly in environments where both human and technical resources are limited. The results highlight not only the feasibility of such an approach but also its potential to significantly enhance the quality, efficiency, and accessibility of digital curatorial work.

Keywords:

Digital cultural heritage, AI curator, GenAI information extraction.