Generative artificial intelligence (GenAI) is reshaping pedagogical practices in higher education, especially in domains such as database management where conceptual modelling and understanding are essential. This paper proposes a shift in teaching strategies to embrace GenAI as a learning tool. We present and discuss a set of diverse exercises considering GenAI existence. We aim to foster students' critical thinking, interpretation skills, and conceptual understanding of topics such as data modeling and SQL query specification in the GenAI era. These exercises have been implemented and preliminarily used in a first-year course on Information Systems at the University of Castilla-La Mancha. Initial results suggest that they could enhance engagement and foster deeper learning. By integrating GenAI into the classroom in a thoughtful and strategic way, educators could promote understanding and prepare students for AI-augmented academic and professional environments dealing with database management.
Decision-making in industrial environments increasingly depends on the ability to align operational data with strategic goals. Data-driven approaches often focus on the available data while overlooking the specific informational needs of decision makers. To bridge this gap, this paper proposes a multi-driven approach that integrates both data-driven and requirements-driven perspectives, supported by the use of Large Language Models (LLMs). Through prompt-based interactions, LLMs generate conceptual and analytical models from metadata and user requirements, accelerating the modeling process while ensuring relevance and coherence through the integration of both data and analytical requirements. The approach is applied to a real-world industrial case characterized by operational complexity, high variability, and technological heterogeneity. The results show how the combined use of LLMs and a structured modeling approach can support the development of analytical systems that are technically feasible and strategically aligned.
Goal-Oriented Requirements Engineering (GORE) supports the development team in identifying the requirements that the system must fulfil. Commonly applied during the initial stages of requirements gathering, it focuses on identifying the system's goals, presenting their decomposition as a means of offering alternatives to satisfy them. Knowledge Acquisition in Automated Specification (KAOS) is a GORE approach that comprises a method, a software environment, and a modeling language. Modelling languages can be adaptable to various domains/application areas where the software will be developed. This way, extensions are proposed to adapt the modelling to the desired scenarios. These adaptations are referred to as extensions. KAOS has been extended to various areas, including security, adaptive systems, and aspects, among others. The creation of new KAOS extensions has been growing. It is expected to continue in the coming years, as it is necessary to adapt languages to the various existing contexts and those that emerge with the constant evolution in software development. Creating an extension is a complex task with inherent challenges, such as maintaining consistency between the developed and existing extensions. Given these facts, we recognise the need to support the creation of KAOS language extensions. This study aims to support the systematic creation of new KAOS extensions through a systematic process. The proposed process was used to create a new KAOS extension to represent accessibility concepts, which proved valid for this purpose. Finally, the PRAOS process was evaluated by KAOS extension specialists through a qualitative study.
The tutorial will introduce a new concept discovery power that can be used to characterize an enterprise modeling language. The concept is different from, but connected to, the concept of expressive power. The concept is defined as "the degree of help provided by the structure of an enterprise modeling language to expand a partly built model or fill gaps in it". The concept is realized by discovery rules that are different for different modeling languages.
While models and digital twins are heavily related, how we use these models, how we understand those models, and even what these models represent for digital twins might change over time. Existing research provides model classifications from different angles, e.g., kind of usage, model content, nature of abstraction; however, a common understanding of which kinds of models could be useful and are used in digital twin engineering and during its lifetime is missing. In this work, we present an initial faceted classification to label models based on different properties relevant for digital twins in manufacturing. We show the application of the classification on different models of a digital twin in manufacturing. The classification can make finding and reusing models in and for digital twins easier, help to identify information gaps, and give inspiration about how models might change their purpose and nature over time.
We introduce a conceptual model for highlights to support automated data analysis and storytelling. Highlights reveal key facts, of high significance, that are hidden in the data with which a data analyst works. The model builds on the concepts of Holistic and Elementary Highlights, along with their context, constituents and interrelationships, whose synergy can identify internal properties, patterns and key facts in a dataset being analyzed. We also report how the related literature fits within the model, as well as a first implementation of it.
Formal ontologies can remedy the ambiguity of narrative financial standards, yet their static nature creates a gap between specification and implementation. This paper presents a framework that transforms static UFO/OntoUML models into interactive, executable artifacts, serving as a validation tool for modelers, a communication bridge for stakeholders, and an executable specification for developers. This framework uses a technology-neutral Event Specification Table (EST) to generate executable logic from ontological patterns, demonstrated with the IFRS 15 standard. Unlike operational platforms, the resulting "living model" is a validation instrument designed to test and communicate the behavioral semantics of complex standards, closing the loop between theory and practice for a more robust development process.
This tutorial offers an introduction to multi-level modeling and corresponding multi-level language architectures. They make it possible to solve serious problems of traditional modeling and programming languages that are limited to two levels. The additional abstraction they provide not only promotes the reusability, integrity and adaptability of languages, models and software systems, but also allows for new software architectures that feature a common representation of models and programs. Users of these systems are empowered to navigate the conceptual foundation of the software they use at runtime, and if needed, adapt it to changing requirements. The multi-level language architecture and corresponding modeling methods that are subject of the tutorial [1, 2] have been developed over the last fifteen years within the project "Language Engineering for Multi-Level Modeling", LE4MM https://www.le4mm.org [2], with roots going back to the ninities of last century.
This study examines the effectiveness of a teaching approach and supporting tools used in a Model-Driven Engineering (MDE) course. Drawing on students' perspectives, it investigates key factors such as ease of use, learnability, satisfaction, perceived complexity, effectiveness, and overall fitness for purpose. With respect to the toolset, the analysis focuses on usability, feedback, and the quality of the learning experience. The course design was informed by best practices in MDE education, combining the MERODE method with the Merlin tool and its prototyper to reduce complexity. A project-based learning strategy was adopted to reflect real-world development contexts. The study involved 36 students from Informatics Engineering and Computer Science programs at two Cuban universities. Data were collected through surveys and analyzed statistically. Results indicate generally positive perceptions of both the course and the tools, while also identifying areas for improvement, particularly in tool learnability and interface design. Beyond technical proficiency, the course aims to instill a deeper understanding of the role of models in software development. These findings may serve as a useful reference for educators seeking to design or enhance MDE courses.
The Digital Transformation of manufacturing, driven by Industry 4.0 and the green transition, has led to a surge in sensor deployment and real-time data collection. This shift gives rise to contexts in which large volumes of complex multivariate time series data present valuable opportunities for advanced analytics, while simultaneously posing significant challenges for effective interpretation. Analytical dashboards are widely used to support decision-making, yet their impact is often limited when design choices do not align with users' analytical goals or cognitive workflows. A promising response to this challenge is data storytelling, which combines data visualization with narrative structures to enhance comprehension, especially in high-pressure, multi-stakeholder environments. However, in complex industrial contexts, the task of identifying and preparing relevant data for analysis presents considerable challenges due to the massive data volume constantly generated. Recent advances in Artificial Intelligence, particularly Large Language Models (LLMs), present new opportunities to automate and enhance the development of such goal-oriented dashboards. It is therefore necessary to investigate how they can be incorporated into a method that applies them for data storytelling in data-intensive contexts. In light of this, this paper proposes a method for designing analytical dashboards that integrate multivariate sensor data with goal-based storytelling techniques, supported by LLMs to accelerate and guide the development process. The proposed method is instantiated in a real-world industrial case, within the PRODUTECH R3 "Industry-UP" project, in the CEI use case for anomaly detection and operational optimization in sensorized stone-cutting machines. The results show that the method reduces manual intervention, identifies data gaps in earlier stages, and delivers dashboards directly traceable to strategic goals, improving both development efficiency and decision-support quality.
Large Language Models (LLMs) have already shown potentially significant capabilities in assisting users with writing and coding tasks. In this paper, we explore how LLM-based assistance can be leveraged in a modeling environment for textual Ontology-Driven Conceptual Modeling. We integrate the UFO-based textual language 'Tonto' with an LLM-powered assistant. We employ detailed UFO-based 'guidance' texts which are included by the modeling environment automatically in the context of user prompts along with the current ontology coding artifacts. The tool can take actions such as creating files, changing code, invoking Tonto syntax verification, while still maintaining the modeler in the loop. Our initial exploration shows that a number of modeling tasks can potentially be automated (such as suggesting new elements, summarizing the model, checking consistency of usage of UFO concepts, model fixing, etc.). The tool is proposed as a testbed for empirical user studies.
The large and heterogeneous data sets that characterize Data-Intensive Domains (DID) pose a challenge to developing data analysis and management approaches. A successful and efficient data-knowledge extraction from DID-based systems is determined by assembling and analyzing such data sets, but integrating their different sources is arduous work. Finding sound solutions for this problem has become a relevant research goal, as existing DID-based systems are not solving it convincingly. To solve this problem, a conceptual characterization of the data sets that constitute DID-based systems is essential. Using foundational ontologies and conceptual modeling provides an adequate strategy to face the complexity of this problem by clarifying the data structure that is to be analyzed and managed. In this project, we tackle this principle by defining a method grounded on a conceptual model to develop efficient DID-based systems and using a well-grounded combination of Explainable Artificial Intelligence (XAI) and Machine Learning (ML) techniques to perform data analytics. In addition, the characterization of a platform for implementing the method has been designed and developed. The project's chosen application domain is genomics, specifically in predicting critical diseases before symptoms manifest. Using XAI and ML with genomic information can contribute to the advancement of precision medicine, allowing the prediction of future diseases based on the available genomic data. The ML dimension covers the predictive knowledge (is a disease present in a patient?), while the XAI dimension deals with the explainable part (why does the patient have a disease?).
Modern healthcare is shifting toward a more personalized approach, where treatments and diagnostics are tailored to the individual. Genetic testing is a key driver of this evolution, offering a powerful way to diagnose and assess health risks based on a person's unique genetic makeup. Traditionally, this analysis has focused on identifying a single, high-impact genetic variant as the primary cause of a patient's symptoms. However, advancements over the last few years have revealed that most diseases are far more complex, arising from the combined influence of numerous variants across the entire genome. This new perspective drives the daily generation of vast and complex data, creating an urgent need for genomic information systems to evolve in support of this new disease paradigm. The PROS Research Group specializes in creating genomic information systems built upon a strong conceptual modeling foundation. The cornerstone of these systems is the Conceptual Schema of the Human Genome (CSHG), which provides a standardized model for genomic data. However, this model remains grounded in the traditional, single-variant perspective. This paper presents an extension to the CSHG that integrates both single-variant and the more complex multi-variant perspective. Ultimately, this contribution provides the foundation for generating trustworthy and transparent information systems that can accurately reflect the full complexity of human disease.
Causal Loop Diagrams (CLDs) document and visualize the dynamics of complex systems, describing their relevant factors, called variables, and causal relationships between them. The most interesting features for exploring CLDs are causal loops, i.e., circuits of causal relationships, which can be characterized as balancing or reinforcing; other interesting features of CLDs are causal routes, i.e., chains of relationships connecting any two pairs of nodes, recognized as increasing or decreasing. We hereby introduce CLD-Explorer, a prototype interactive system for inputting new CLD diagrams and analysing their features, reviewing their causal loops and routes. The system supports simple interfaces for systematically extracting and exploring loops and routes, which can be used by systemic designers in identifying the "areas of intervention" within a complex system, empowering organized reasoning on CLDs and improving the related decision-making processes.
Recent advances in AI allow decision makers to move beyond traditional analysis towards sophisticated decision-making tasks that require human intuition and perception. This opens the door to a novel form of OLAP, one that integrates unstructured data -such as text and images- into its analytical workflows. In this work we propose OLAP-AI, a novel and enriched form of the OLAP paradigm aimed at supporting, besides the analysis of categorical and numeric data, also semantically rich operations over free text and images. OLAP-AI is multi-modal, in that it supports crossed semantic searches between text and images for filtering and grouping. Besides, it significantly extends aggregation by operating on text and images rather than on numeric data only, and by relying on generative models. To investigate the technical feasibility of the OLAP-AI paradigm, we provide a proof-of-concept by relying on the open-source vector DBMS Weaviate.
Virtual Reality (VR) is increasingly employed in applications such as virtual exhibitions, safety training, and medical simulations. Understanding user behavior in these environments is essential for enhancing user experience and informing interaction design. However, analyzing VR behavior is complex due to the multimodal nature of interactions- encompassing gestures, gaze, and speech- which results in vast and heterogeneous data streams. Effectively managing and interpreting this data remains a significant challenge. Well-founded ontologies offer a structured approach to understanding, organizing, and analyzing behavioral data, yet their application in VR behavioral research remains underexplored. To address this gap, we introduce OnBehaVR-an ontology designed to represent and analyze user behavior in VR. Developed using the Unified Foundational Ontology ( UFO) and the OntoUML conceptual modeling language, OnBehaVR provides a formal framework for conceptualizing VR interactions. It aims to (1) clarify the conceptualization of user behavior in VR, (2) enable data integration within this domain, and (3) facilitate behavior analysis through automated reasoning and semantic queries. By leveraging ontology-driven approaches, OnBehaVR contributes to the systematic study of multimodal interactions and supports the development of adaptive and personalized virtual experiences.
This tutorial shows how traditional Entity/Relationship modeling and modern graph data modeling can be combined to bring forward well-designed graph data models that process workloads and maintain data integrity efficiently.
This poster paper aims to aid in bridging the gap between funding organizations, incl. industry partners, and PhD Program providers, allowing for a more transparent and value-creating collaboration between both parties and the PhD Candidates. It introduces the initial version of domain ontology of a PhD program as an entry point for development of a more robust future version of a broader domain ontology of PhD studies, which will serve as an input for unpacking of both PhD Program's and PhD Candidate's funding needs. The domain ontology is modelled using OntoUML modelling language, which anchors this work in the UFO foundational ontology. This enables to capture the human understanding of this domain, which will help to articulate domain specific nuances to all parties involved, for purposes of analysis and clear communication of each party's needs, expectations and agreed outcomes.
The usage of personas, fictional representations of user archetypes, during the initial stages of requirements elicitation has yielded significant improvements in mitigating assumption bias, stemming from the development engineers' preconceptions about end users that frequently do not accurately reflect the actual needs of those users. This paper explores how personas can be used to support automated testing workflows, bridging the gap between higher-level conceptualizations and code. Although personas are commonly used in requirements engineering and during software development processes, they are rarely used for model-based testing. Our approach uses personas in automated testing to construct a software solution that is capable of validating system requirements captured in personas. We propose a domain-specific language capable of encompassing the high-level requirements of a target system through scenarios, flows, and execution plans, and we generate the test suite infrastructure, which features test scenario skeletons. Although these skeletons require handwritten code to perform the execution steps, this process may be further automated to produce executable source code with minimal human intervention.
Conceptual modeling techniques serve as a foundation for reliable process representations across various domains. In business process modeling, challenges such as incomplete, outdated, or entirely absent process documentation lead to reliance on extensive communication between domain experts and modelers. Business Process Modeling and Notation (BPMN) has become a standard in addressing these challenges, yet generating accurate BPMN models from textual descriptions remains difficult. This paper introduces LLM4BPMNGen, a tool leveraging LLMs to directly transform textual process descriptions into BPMN XML models, and covering a large set of BPMN elements. Preliminary analysis indicates high usability, although minor layout adjustments were suggested. Future work aims to address performance issues of the tool and enhance flexibility regarding LLM providers.
