Hybrid Event

Event program

Wednesday, 6/4/2025 4:00 PM - 7:00 PM, Leut II, Hotel Admiral, Opatija 4:00 PM - 4:30 PM Invited Lecture  Matteo Francia (University of Bologna, Cesena, Italy) Unified Data Platforms: Challenges and Opportunities in the Era of Heterogeneous Big Data  4:30 PM - 7:00 PM Papers  1. A. Meštrović, S. Beliga (Sveučilište u Rijeci, Fakultet informatike i digitalnih tehnologija, Rijeka, Croatia), D. Pitoski (Sveučilište u Rijeci, Centar za umjetnu inteligenciju i kibernetičku sigurnost, Rijeka, Croatia) Peoplet: Exploring Organizational Structures through Social Network Analysis  In this paper, we propose a conceptual model based on multilayer networks that enables representing various aspects of relationships within the organizational structure. The model captures multiple layers of organizational interactions, including hierarchical relationships, connections based on shared expertise, common interests, and participation in joint projects. A distinctive feature of our approach is the integration of survey data for constructing specific network layers, enabling a more detailed representation of interpersonal and intergroup dynamics. Surveys allow for the direct capture of certain relationships, such as shared interests or informal connections, which may not be evident from other data sources. This modality ensures that the constructed network reflects both formal structures and underlying relational dynamics. Furthermore, we define a systematic approach for the visualization, and analysis of organizational networks. This approach is implemented within the Peoplet application, which enables detailed analysis of organizational networks at global, meso, and local levels to reveal patterns and insights across different relational dimensions. Our findings demonstrate how this multilayer approach provides a richer understanding of organizational structures and dynamics, offering valuable perspectives for optimizing collaboration, communication, and decision-making within organizations. 2. I. Pavić (University of Zagreb, Faculty of electrical engineering and computing, Zagreb, Croatia), F. Varga (University of Zagreb, Faculty of agriculture, Zagreb, Croatia), I. Čavrak (University of Zagreb, Faculty of electrical engineering and computing, Zagreb, Croatia) Assessing Spatial Accuracy of Nature Observations Using the LLM-Based Geoparser Service  Geo-parsing tools have historically relied on AI-based techniques for tasks such as text processing and toponym disambiguation. However, the rapid proliferation of large language model based tools and services necessitates a re-evaluation of traditional geoparser architectures to explore performance enhancements, introduce new functionalities, support different languages, or allow for rapid prototyping and development of new geoparser-based services. This paper focuses on the functional components of typical geoparsers and examines the effects of substituting one or more of these elements with LLM-based functionality. The performance impact of these modifications is evaluated in the context of a developed online service for assessing the spatial accuracy of nature observations. The performance of the service is compared against the baseline performance of the open-source geoparser, Mordecai3, using datasets in Croatian language and with varying complexity of location descriptions. 3. I. Mekterović, L. Brkić (Faculty of Electrical Engineering and Computing, Zagreb, Croatia) Can LLM Pass a CS1 Course?  ChatGPT and other Large Language Models (LLMs) have significantly impacted a number of fields by offering sophisticated natural language processing capabilities. LLMs produce remarkable outcomes in higher education, particularly in introductory computer science courses (CS1). This study examines the changing performance of two cutting-edge LLMs, ChatGPT and Gemini, in the setting of a university-level introductory computer science course (CS1). These LLMs were 'enrolled' as students for two consecutive academic years and given a wide range of assignments: lab exercises, midterm exam, and final exam. The fact that only one of the two LLMs finished the course in the first year showed the limitations of these models at that stage. But there was a noticeable change in the second year, when both LLMs received passing ratings, indicating the quick development of LLMs. This study highlights how LLMs can revolutionize computer science higher education by creating chances for individualized learning, feedback and advanced teaching strategies. It also calls into question the authenticity of student work, academic integrity, and the necessity for teachers to modify their pedagogical approaches to successfully incorporate and utilize these powerful technologies while upholding the fundamental principles of their courses. 4. A. Rajh (Faculty of Humanities and Social Sciences, Zagreb, Croatia) Semantic and Other Serializations of Archival Description: Impact on the Understanding, Processing, and Use of Archival Materials  Archival science is a branch of information sciences that focuses on the information contained in archival materials, the materials themselves as a form of this information, and the metadata inscribed in finding aids. Archival description is a crucial responsibility of archivists. This responsibility depends on descriptive standards to guarantee global understanding among them. The framework was established in the 1990s and 2000s with the General International Standard Archival Description – ISAD(G), related standards, and software-readable finding aids in XML serialization. The most recent standard, Records in Contexts (RiC), is based on Semantic Web concepts, enabling more interoperable machine processing through a shared ontology. It also facilitates automated reasoning and new uses that may not need to be anticipated by archivists creating descriptions. RiC depends on serializing archival descriptions in RDF/XML and Turtle. The author of this paper explores the creation and validation of XML, Turtle, and Notation3 serializations for archival purposes, focusing on their features, ease of machine processing, readability, and suitability for formulating queries and extracting information from archival materials. The author concludes by discussing the implications of different serializations of descriptions and recommends enhancing RiC's reasoning capabilities to support a broader range of archival information applications. 5. D. Begušić (Faculty of Electrical Engineering and Computing, Zagreb, Croatia), D. Pintar (FAKULTET ELEKTROTEHNIKE I RAČUNARSTVA, ZAGREB, Croatia), Z. Smole (Modalis Ltd, Zagreb, Croatia), B. Dalbelo Bašić (Faculty of Electrical Engineering and Computing, Zagreb, Croatia) Large-scale Molecular Similarity Search using Vector Database  Molecular similarity search is a task in the field of cheminformatics that enables researchers to identify molecules with structural or chemical properties similar to a target compound. This study demonstrates an approach of leveraging vector databases for more efficient large-scale molecular similarity search to facilitate fast comparisons using advanced database indexing techniques. Molecules are often represented as high-dimensional vectors that aim to capture their essential chemical features. That makes for a logical application of the vector databases in efficient molecular similarity search. In comparison with traditional similarity search data architecture, this method significantly reduces computational costs by avoiding the prohibitive expense of calculating complex molecular similarities in real-time and by using advanced clustering methods making large-scale molecular databases more accessible for applications in drug discovery. In this paper we implement a vector database for storing molecular data. Our implementation showcases the current status of the vector database technology for cheminformatics applications and provides a solution for storing and efficient handling of vast openly available molecular datasets. The implemented functionality of molecular similarity search displays greater efficiency compared to currently available solutions. 6. S. Delalić (Faculty of Science, University of Sarajevo & Info Studio d.o.o., Sarajevo, Bosnia and Herzegovina), S. Behić (Faculty of Electrical Engineering, University of Sarajevo & Info Studio d.o.o., Sarajevo, Bosnia and Herzegovina), H. Goralija (Info Studio d.o.o., Sarajevo, Bosnia and Herzegovina) Enhancing Warehouse Efficiency: Estimating Order Picking Complexity for Improved Performance  Warehouse Management Systems (WMS) employ advanced optimization techniques to enhance efficiency and streamline processes, from inventory positioning to order picking and packing. Among these, order picking represents the most time-consuming and resource-intensive operation. This paper presents a novel approach for monitoring worker efficiency in warehouses, focusing on estimating the complexity and time required for order picking. A variety of factors influence these estimates, including item location, quantity, dimensions and weight of items, picking sequence, and whether the location is in the stock or picking zone. Accurate estimation enables effective daily work planning, real-time monitoring of worker productivity, and overall warehouse efficiency. The proposed approach has been tested in real-world warehouse environments, demonstrating its practical applicability and potential to significantly improve worker performance, resource allocation, and operational management. 7. L. Brkić, I. Mekterović, M. Fertalj (University of Zagreb Faculty of Electrical Engineering and Computing, Zagreb, Croatia) Enhancing Accuracy in Peer Assessment by Addressing Bias and Outliers through Calibration Method  Peer assessment is increasingly used in educational settings as a tool to enhance learning quality and critical thinking. However, this process comes with challenges, such as ensuring the fairness and accuracy of evaluations. In this article, we explore methods for analyzing and correcting student grades within peer assessment systems. Special emphasis is placed on identifying and quantifying bias—tendencies of individual students to systematically assign higher or lower grades—and on detecting and managing outliers—grades that significantly deviate from the average. We also examine the calibration method, in which students evaluate tests graded by instructors, allowing for the assessment of individual evaluator accuracy. By combining these methods, the accuracy and reliability of peer assessment can be improved, and the results of this research provide guidelines for designing more effective assessment systems. 8. L. Štefan (University of Zagreb Faculty of Civil Engineering, Zagreb, Croatia), L. Đud, D. Pintar (University of Zagreb Faculty of Electrical Engineering and Computing, Zagreb, Croatia), K. Potočki (University of Zagreb Faculty of Civil Engineering, Zagreb, Croatia) Overview of Machine Learning Methods for Analyzing and Clustering of Flood Wave Data  The design hydrograph, which incorporates peak flow, duration, and volume of flood waves, is essential for designing hydraulic structures and flood protection systems. In many cases, the shape of the hydrograph, in addition to its peak value, plays a significant role in accurately representing flood behavior. The paper provides an overview of various machine learning techniques for hydrograph classification and clustering, including unsupervised cluster analysis, K-means clustering, fuzzy c-means, and parameter-based classification. Additionally, it presents advanced methods such as bivariate copula analysis and functional data representation for modeling flood wave shapes. These approaches provide a more flexible and data-driven framework for analyzing flood events, offering advantages over traditional methods. As an example, the paper presents results from K-means shape-based clustering of flood waves represented by Bspline functions on Kupa and Krapina River catchments. The paper also highlights the potential for applying these techniques to different catchments and their broader applicability in hydraulic design.

Basic information:

Boris Vrdoljak (Croatia), Matteo Golfarelli (Italy), Mihaela Vranić (Croatia)

Marko Banek (Croatia), Matteo Golfarelli (Italy), Damir Pintar (Croatia), Mihaela Vranić (Croatia), Boris Vrdoljak (Croatia)

Mirta Baranović (Croatia), Ladjel Bellatreche (France), Ljiljana Brkić (Croatia), Alfredo Cuzzocrea (Italy), Matteo Francia (Italy), Enrico Gallinucci (Italy), Paolo Garza (Italy), Marko Gulić (Croatia), Luka Humski (Croatia), Igor Mekterović (Croatia), Sandro Skansi (Croatia), A Min Tjoa (Austria)

Registration / Fees:

The student discount doesn't apply to PhD students.

NOTE FOR AUTHORS: In order to have your paper published, it is required that you pay at least one registration fee for each paper. Authors of 2 or more papers are entitled to a 10% discount.

Boris Vrdoljak
University of Zagreb
Faculty of Electrical Engineering and Computing
Unska 3
HR-10000 Zagreb, Croatia

Phone: +385 1 6129 532
Fax: +385 1 6129 915
E-mail: boris.vrdoljak@fer.hr

The best papers will get a special award.
Accepted papers will be published in the ISSN registered conference proceedings. Papers in English presented at the conference will be submitted for inclusion in the IEEE Xplore Digital Library.