Keywords: Conceptual modeling, Data modeling, ER, Genetic algorithms, Graph neural networks, Model Modularization
Astract: Due to the crucial role conceptual models play in explicitly representing a subject domain, it is imperative that they are comprehensible and maintainable by humans. Modularization, i.e., decomposing an overarching, monolith model into smaller modules, is an established technique to make the model comprehensible and maintainable. Genetic Algorithms (GA) have been applied to modularize conceptual models by formulating desired structural model characteristics as multiple objectives. Recently, Graph Neural Networks (GNN)-based methods have shown promising performance in graph processing tasks, including graph clustering but outside the conceptual modeling domain. In this paper, we present a novel approach for GNN-based conceptual model modularization and comparatively analyze our approach against an existing multi-objective GA-based one. Furthermore, we provide a comparative analysis of our novel GNN model against two existing GNN-based graph clustering approaches. We investigate the dependence of the quality of the modularized solutions on the model size. We discuss the comparative results of our novel GNN-based approach and the existing GA-based approach to derive future research lines. Furthermore, our results show, that our proposed GNN-based modularization outperforms the existing GNN-based graph clustering approaches and provides a suitable alternative compared to the GA-based modularization.

Keywords: Conceptual modeling, Domain modeling, Generative AI, LLM, UML
Astract: The information systems engineering community is increasingly exploring the use of Large Language Models (LLMs) for a variety of tasks, including domain modeling, business process modeling, software modeling, and systems modeling. However, most existing research remains exploratory and lacks a systematic approach to analyzing the impact of prompt content on model quality. This paper seeks to fill this gap by investigating how different levels of description granularity (whole text vs. paragraph-by-paragraph) and modeling strategies (model-based vs. list-based) affect the quality of LLM-generated domain models. Specifically, we conducted an experiment with two state-of-the-art LLMs (GPT-4o and Llama-3.1-70b-versatile) on tasks involving use case and class modeling. Our results reveal challenges that extend beyond the chosen granularity, strategy, and LLM, emphasizing the importance of human modelers not only in crafting effective prompts but also in identifying and addressing critical aspects of LLM-generated models that require refinement and correction.

Keywords: Artificial intelligence, Cloud platform, Conceptual modeling, Knowledge graph, Model transformation, Model-driven engineering
Astract: Semantic processing of conceptual models is a focus of research for several years, bridging the disciplines of knowledge-based systems, conceptual modeling, and model-driven software engineering. With Knowledge Graphs, this research area gained momentum. In this paper, we introduce CM2KGcloud, a generic and extensible Web-based platform for transforming conceptual models into Knowledge Graphs. The platform can work on models created by state-of-the-art metamodeling platforms (e.g., EMF, Papyrus, ADOxx) and transforms models created with them into standardized Knowledge Graph representations like GraphML, RDF, and OWL. CM2KGcloud can be used as a service and can be integrated into software systems by its exposed API.

Keywords: Graph Neural Networks, Ontology-Driven Conceptual Models, Pre-trained Language Model, Representation Learning
Astract: Conceptual models (CMs) offer a structured way to organize and communicate information in information systems. However, current models lack adequate semantics of the terminology of the underlying domain model, leading to inconsistent interpretations and uses of information. Ontology-driven conceptual modeling languages provide primitives for articulating these domain notions based on the ontological categories, i.e., stereotypes put forth by upper-level (or foundational) ontologies. Existing CMs have been created using ontologically-neutral languages (e.g., UML, ER). Enriching these models with ontological categories can better support model evaluation, meaning negotiation, semantic interoperability, and complexity management. However, manual stereotyping is prohibitive, given the sheer size of the legacy base of ontologically-neutral models. In this paper, we present a graph language modeling framework for conceptual models that combines finetuning pre-trained language models to learn the vector representation of OntoUML models’ data and then perform a graph neural networks-based node classification that exploits the model’s graph structure to predict the stereotype of model classes and relations. We show with an extensive comparative evaluation that our approach significantly outperforms existing stereotype prediction approaches.

Keywords: Domain Modeling, Large Language Model, Process Mining, UML
Astract: Large Language Models (LLMs) have opened new opportunities in modeling in general, and conceptual modeling in particular. With their advanced reasoning capabilities, accessible through natural language interfaces, LLMs enable humans to deepen their understanding of different application domains and enhance their modeling skills. However, the open-ended nature of these interfaces results in diverse interaction behaviors, which may also affect the perceived usefulness of LLM-assisted conceptual modeling. Existing works focus on various quality metrics of LLM outcomes, yet limited attention is given to how users interact with LLMs for such modeling tasks. To address this gap, we present the design and findings of an empirical study conducted with information systems students. After labeling the interactions according to their intentions (e.g., Create Model, Discuss, or Present), and representing them as an event log, we applied process mining techniques to discover process models. These models vividly capture the interaction behaviors and reveal recurrent patterns. We explored the differences in interacting with two LLMs (GPT 4.0 and Code Llama) for two modeling tasks (use case and domain modeling) across three application domains. Additionally, we analyzed user perceptions regarding the usefulness and ease of use of LLM-assisted conceptual modeling.