Parallel Session 6 – Artificial Intelligence for/in/and Research Libraries II
Date: Thursday, 3 July 2025, from 09:00 to 10:30
Moderator: Marion Brunetti, Enssib, France
Location: Room 1129
6.1) AI-Driven Approaches for Enhancing Text Understanding and Information Extraction from Historical Newspapers in Serbia
Presenter: Aleksandra Trtovac, University Library Belgrade, Serbia
The development of advanced language models and technologies can significantly improve the accuracy of information retrieval from scanned newspaper pages and their integration with online knowledge bases. This is particularly crucial for so-called “small languages”, which must develop their own language resources. As artificial intelligence (AI) gains prominence in digital humanities, library science, and archival research, there is an increasing need for the development of Archival and Open Linked Data. AI offers great potential to enhance digital archives by expanding linked data and innovating access methods. Automating these processes can reduce resource requirements, improve efficiency, and support expert work. However, it is essential to consider the limitations, biases, and ethical implications of such technologies.
The Searchable Digital Library at the University Library Belgrade (www.pretraziva.rs) is a key resource, comprising over 1.3 million pages of historical newspapers, literary magazines, and special collection monographs. To modernize the approach to this corpus, the new research draws on the activities of earlier, successfully implemented projects related to distant reading, such as COST CA16204 (Distant Reading for European Literary History). This paper also outlines innovations aimed at improving the semantic visibility of the collection of historical newspapers, supported by the Ministry of Culture of the Republic of Serbia, in the framework of the project Education of the library and research community on the model of increasing the visibility of Serbian historical newspapers through advanced language models and technologies. These innovations, which can be applied to any “small language”, are continuously implemented by the University Library Belgrade.
Key innovations, that are still being worked on, include:
1. Enhanced Search Capabilities: Expanding search queries to account for grammatical forms such as cases, utilizing web services from the Society for Language Resources and Technologies and Serbian electronic dictionaries.
2. Information Extraction: Automatically identifying, tagging, and extracting key information such as article titles, dates, names of people, organizations, and locations.
3. Named Entity Recognition and Linking: Identifying significant entities and their linking using semantic networks like WordNet, Wikidata, and GeoNames, along with word embeddings developed within the TESLA (Text Embeddings – Serbian Language Applications) project.
4. Geographic Mapping: Using recognized and linked location data to map places mentioned in articles, linking geographic locations to specific events or time periods.
5. Interactive Visualizations: Developing interactive visual tools (graphs, maps, etc.) that enable users to explore and understand historical information from newspaper articles.
The innovations presented aim to improve text search and information extraction from scanned newspapers, create visualizations for historical research, enhance search and recommendation features, and support SDG principles through the quality education of librarians, researchers, students, seniors, and other end users.
These developments hold promise for better accessibility to historical data, enabling more effective research and discovery. By combining AI with Archival Linked Data, researchers can uncover new insights from historical texts while addressing challenges related to “small languages” and resource constraints. This approach ensures that the work is not only practical but also mindful of the ethical dimensions involved in using AI for archival purposes.
6.2) Artificial Intelligence, Real Library: Insights from Project Laibro for Organizational Development
Presenter: Leticia Antunes Nogueira, NTNU – Norwegian University of Science and Technology, Norway
This presentation reports the results of Project Laibro: artificial intelligence, genuine library and discusses implications for organizational development from a leadership perspective.
Project Laibro was initiated by NTNU library (Norwegian University of Science and Technology) in November 2023, with the purpose of ensuring that the library continues to be a valuable and adaptable resource in the university ecosystem, as well as an active contributor to the development of research libraries — particularly regarding technological change and artificial intelligence. Project Laibro’s main objective was to establish the foundations for the library’s AI strategy, which is divided into three sub-goals: (1) to establish the necessary knowledge base concerning AI in research libraries; (2) to provide a unified perspective on the topic at NTNU’s library; and (3) to give the library’s leadership recommendations for further action. The project’s final report was delivered in December 2024.
Based on a mix of an employee survey, a survey with library users, interviews with users and with other academic libraries, as well as analysis of the service portfolio of the library in light of the possibilities and challenges brought about by AI, Project Laibro has offered in-depth knowledge for the development of competences and services in academic libraries. The project has analysed the driving forces behind AI development, with a focus on the implications for academic libraries, as well as for research and higher education more broadly. The project’s comprehensive report addressed various themes of relevance, including Copyright and intellectual property, consequences for academic publishing models and agreements, the use of AI in information retrieval and literature search, and citation practices when it comes to AI.
Expanding from the knowledge and experiences accumulated through Project Laibro, the presentation focuses on the Project teams ten recommendations for NTNU Library’s further work with AI, as well as on the process of organizational development sparked by the project. Together, the project leader and the library director offer a leader’s perspective on what these findings mean for research libraries looking forward. Through the experiences at NTNU, we offer the audience concrete insights that are grounded in systematic exploration and organizational development. The ambition is that these insights can add value to library communities in Europe and beyond, and support them as engaged and trusted knowledge hubs, as envisioned in LIBER’s strategy for 2027.
6.3) Applying Decision-Making Models in Library Leadership in Times of AI
Presenters: Liisi Lembinen, University of Tartu, Estonia; Heli Kautonen, University of Turku, Finland
The rapid emergence of large language models (LLMs) and other AI tools in academia presents significant challenges for library directors, necessitating strategic and informed decision-making. Library patrons and staff are seeking guidance on navigating the current technological landscape and preparing for future advancements. This situation raises the critical question: How should the library leadership tackle integrating new AI technologies in their organisation?
Recent research indicates that directors from LIBER member libraries employ various decision-making models depending on the situation. The decision-making process varies significantly between long-term strategic planning, immediate crisis management, and innovation decisions. Traditional decision-making frameworks, such as the Vroom-Yetton-Jago model and the Tannenbaum-Schmidt Leadership Continuum, have been effectively utilized in library leadership. However, the advent of AI introduces new dynamics that may alter these established processes.
The AI phenomenon can be examined from three perspectives: innovation, long-term strategy, and crisis management. Each perspective requires distinct decision-making approaches, yet the rapid development of AI tools often blurs these boundaries, creating a complex environment for library leaders.
In this presentation, we explore how library directors can apply established decision-making models to address the challenges posed by AI. We will draw on examples from existing literature and practical experiences, including case studies from Turku University Library, to illustrate the practical implications of these models in real-world scenarios. Our analysis aims to provide actionable insights for library leaders to effectively manage AI integration and lead their institutions through this transformative period. It also looks at how AI could possibly enhance the future of decision-making.