Keywords: Model-Driven Engineering, Artificial Intelligence

Keywords: AI Interpretability, Embedding Space, Internal Representations, Multimodal Data, Semantic Event Matching
Astract: Traditional process mining relies on symbolic event logs that represent activities as discrete labels, often overlooking the rich contextual and semantic nuances found in real-world data such as textual reports, visual records, or sensor outputs. In this paper, we propose a paradigm shift: using the internal representations of AI models—embedding spaces learned from data—as the foundation for process mining. Our framework performs both process discovery and conformance checking directly in these continuous vector spaces, enabling the detection of semantically similar yet lexically divergent events. We evaluate our approach along three dimensions: (i) whether embedding-based discovery maintains or improves accuracy over symbolic baselines, (ii) whether multimodal sources such as video and audio can be processed as unified embeddings for mining purposes, and (iii) whether conformance checking in embedding space enables alignment across noisy or semantically perturbed traces. By treating AI’s internal representations as a novel form of process evidence, we show how process mining can move beyond traditional logs and unlock deeper, semantically enriched interpretations of real-world workflows.

Keywords: Business process simulation, Video generation, Process modeling
Astract: Video generation models have opened new opportunities for simulating business processes through realistic visualizations. However, current video generation approaches often fall short of capturing the inherent dynamics and structure of business processes and tend to produce inconsistent simulations that lack the rigor provided by formal process models. To address these limitations, we introduce a novel method termed Petri Net structure-driven video generation, which integrates the inherent structural information from process models to tailor video simulations more closely to the dynamics of business processes. We explore multiple strategies for this tailoring, including i) the use of domain knowledge-rich prompting, ii) a storyboard employing image references extracted from process evidence data, and iii) generated image references informed by process models. We evaluate our method across diverse domains, and demonstrate that the Petri Net structure-driven approach improves the perceived usefulness and consistency of the simulated video, marking a step forward in the use of generative AI for more realistic business process simulation.

Keywords: API, LLM, Microservices
Astract: Linking APIs to the business functions they implement is crucial for handling software operations, especially during recovery from disasters or outages. In this context, the speed and accuracy of operators in linking them impact response time during mission-critical operation activities. Besides, this linkage is essential to designing preventive actions, such as resilience strategies. Automatic API classification using Large Language Models (LLMs) may simplify and speed up APIs-business function linkage. However, previous studies unveiled the barriers practitioners face when deciding on and adopting LLMs in software engineering (SE) tasks due to a lack of guidance for non-experts. This paper aims to lower barriers to using LLMs by systems operators and site reliability engineers (SREs), focusing on the API classification task in the context of operational activities. Based on three cases from the finance industry, we extracted requirements for LLM usage, and assessed 14 recently released LLMs on this task. Our results demonstrate that LLMs accurately classify APIs using business function targets with an F1–Score of 89.5 for the leading LLM without requiring specific LLM expertise and resource-intensive fine-tuning. Besides, our findings on LLMs’ performance and reliability mark a significant advancement in comparing open and closed-source and general and domain-specific LLMs in an SE classification task. Eventually, our experiments yield practical guidance for implementing LLMs in this context. Artifacts used in and generated by the experiments are publicly available at https://bit.ly/llms4apiclassification.

Keywords: Model repository, Dataset, UML, Open models, Curation, Community, Education, Machine learning
Astract: Datasets of Unified Modeling Language (UML) models are becoming increasingly valuable for education, empirical research, and tool development in model-driven engineering (MDE) and conceptual modeling. In recent years, several datasets have emerged - mostly compiled through automated crawling of open platforms such as GitHub and GenMyModel. While these efforts have improved access to real-world modeling artifacts, the resulting collections often suffer from serious quality issues: they include syntactically invalid models, semantically incorrect structures, and placeholder or dummy content. Moreover, most models are not accompanied by textual domain descriptions, which are essential for understanding the intent behind the model and assessing its semantic soundness. Therefore these model datasets are far from ideal as a source for modeling exercises or empirical MDE research. This paper presents an initial step toward a community-curated golden dataset of UML models, designed to address these limitations. Our contribution includes i) a curated set of UML models, each paired with a natural language description of the modeled domain requirements, ii) a publicly accessible web platform for exploring and querying the dataset, and iii) a structured process for community-based contribution and evaluation to support sustainable growth and quality assurance of the dataset. By fostering community involvement and providing high-quality, semantically grounded models, this work lays the foundation for a widely accepted benchmark dataset in UML-based research and education.