New Scientific Paper by the Forensic Science Centre Ivan Vučetić on the Use of Artificial Intelligence in Forensic Authorship Analysis

A scientific paper titled “The Small Sample Challenge: Unsupervised Authorship Clustering in Croatian” has been published in Inventions (MDPI), an international scientific journal indexed in the Web of Science and Scopus databases, with an impact factor (IF) of 2.4. One of its co-authors is Andrea Ledić, PhD, Head of the Forensic Science Centre Ivan Vučetić.

The authors are Associate Professor Lidija Tepeš Golubić, PhD; Ian Marković; Andrea Ledić, PhD; and Brigitta Cafuta. The research is the result of interdisciplinary collaboration between the Forensic Science Centre Ivan Vučetić, Zagreb University of Applied Sciences, and the Faculty of Electrical Engineering and Computing, University of Zagreb.

The research addresses a significant challenge in forensic linguistics: identifying possible common authorship of anonymous and threatening texts when only a limited number of samples are available for analysis. Small datasets are common in real-world forensic casework, making the application of conventional analytical methods and contemporary computational approaches a particular methodological challenge.

The study applies unsupervised machine learning methods to group texts according to their linguistic features, without prior knowledge of authorship. Using a corpus of threatening letters written in Croatian, the study examines the potential to identify patterns that may indicate a common author and to develop a more effective model for this type of analysis.

The research is particularly valuable because it combines forensic linguistics, computer science, and artificial intelligence methods. This interdisciplinary approach opens up opportunities to develop new tools that could provide experts with additional objective support when analysing questioned and anonymous texts and contribute to the further development of methods applicable in forensic practice.

The Centre’s contributions included conceptualization, validation, provision of research resources, and review and editing of the manuscript. This brought a forensic expert perspective directly into the development and evaluation of the research approach.

The research was conducted as part of the scientific project “Application of VR technology and neural networks in the field of computer security and digital forensics”, funded by the European Union through the NextGenerationEU Recovery and Resilience Facility, with funding allocated through the Ministry of Science, Education and Youth.

The publication of this paper further demonstrates the Forensic Science Centre Ivan Vučetić’s commitment to developing forensic methods on a scientific basis, applying new technologies, and collaborating with the academic community to connect scientific research with the needs of contemporary forensic practice.

The paper was published on 8 September 2026 in Inventions, Volume 11, Issue 5, Article 93.
Tepeš Golubić, L., Marković, I., Ledić, A., & Cafuta, B. (2026). The Small Sample Challenge: Unsupervised Authorship Clustering in Croatian. Inventions, 11(5), 93. DOI: 10.3390/inventions11050093.

The full scientific paper is available here The Small Sample Challenge: Unsupervised Authorship Clustering in Croatian
 

Page