KEOD 2025 Abstracts


Full Papers
Paper Nr: 41
Title:

From Risk Storylines to a Risk-Driven Ontology of Urban Systems

Authors:

Cristine Griffo, Liz Jessica Olaya Calderon, Massimiliano Pittore and Theodore G. Shepherd

Abstract: Ontological models empower stakeholders to establish shared ontological commitments for achieving objectives, including (1) fostering domain-specific understanding; (2) formalizing communication between stakeholders and modelers; and (3) enabling knowledge inference through formal rule-based systems. A significant challenge arises as conceptual modeling transitions from single-organizational contexts to heterogeneous, multi-perspective environments, raising questions about how quasi-universal conceptualizations can ensure data interoperability. To address this, we propose storylines to integrate diverse perspectives across past and future scenario narratives. This study applies risk-oriented storylines and ontologies through a middle-out approach, synthesizing top-down and bottom-up strategies, in the ontology engineering of urban systems at risk. The results demonstrate that storylines effectively surface domain-specific terminology among stakeholders but exhibit limitations in capturing abstract, generic concepts and relationships. Conversely, the top-down approach (guided by competency questions, literature, and interviews) revealed imperceptible abstract concepts that storylines overlooked, while missing specialized terms identified through narrative methods. These results highlight the complementary value of hybrid methodological frameworks: the middle-out approach mitigates blind spots inherent to purely top-down or bottom-up strategies, enabling more robust ontology development in complex, multi-stakeholder environments. This work advances pragmatic methodologies for interoperable ontology design in urban systems, with implications for risk management and urban resilience planning.
Download

Paper Nr: 77
Title:

The Semantic SI Ontology: Engineering, Alignment, and Validation of a Semantic SI Model

Authors:

Moritz Jordan, Giacomo Lanza, Shanna Schönhals and Sören Auer

Abstract: The Semantic SI Ontology (SIS) provides a semantic model for representing the value of a physical quantity, including kind of quantity, International System of Units (SI) measurement unit, and measurement uncertainty, in accordance with the official Bureau International des Poid et Mesures (BIPM) recommendations. Developed as a formal counterpart to the Digital System of Units XML schema definition (D-SI XSD), the ontology enables harmonized, machine-readable, and interoperable representation of metrological data. This paper outlines the design rationale, the transformation methodology from eXtensible Markup Language (XML) Schema to Web Ontology Language (OWL), and its alignment with existing semantic standards such as Quantities, Units, Dimensions, and Data Types Ontologies (QUDT) and the SI Reference Point. Core ontology structures - such as QuantityValue and MeasurementUncertainty - are discussed in detail. The paper also presents the validation framework, including a Python toolchain for generating OWL individuals from XML, Shapes Constraint Language (SHACL)-based shape validation, and reasoning with established OWL reasoners. Applications in ongoing projects, such as the Metadata4Ing and the Digital Calibration Certificate (DCC) ontology, demonstrate the practical relevance of the SIS as a foundational component in the digital transformation of metrology.
Download

Paper Nr: 88
Title:

Comparative Analysis of Entity Matching Approaches for Product Taxonomy Integration

Authors:

Michel Hagenah and Michaela Kümpel

Abstract: This work examines different approaches to solving the entity matching problem for product categories by converting the GS1 Global Product Categorization (GPC) published by GS1 as an ontology and linking it to the Product Knowledge Graph (ProductKG). For the implementation, methods were developed in Python for word embeddings, WordNet, lemmatization, and large language models (LLMs), which then link classes of the GPC ontology with the classes of the ProductKG. All approaches were carried out on the same source data and each provided an independent version of the linked GPC ontology. As part of the evaluation, the quantities of linked class pairs were analyzed and precision, recall, and F1 score for the Food / Breakfast segment of the GS1 GPC taxonomy were calculated. The results show that no single approach is universally superior. LLMs achieved the highest F1-score due to their deep semantic understanding but suffered from lower precision, making them suitable for applications requiring broad coverage. Lemmatization achieved perfect precision, making it ideal for use cases where false matches must be avoided, though at the cost of significantly lower recall. WordNet offered a balanced trade-off between precision and recall, making it a reasonable default choice. Word embeddings, however, performed poorly in both metrics and did not outperform the other methods.
Download

Paper Nr: 94
Title:

Toward Semantic Explainable AI in Livestock: MoonCAB Enrichment for O-XAI to Sheep BCS Prediction

Authors:

Nourelhouda Hammouda, Mariem Mahfoudh and Khouloud Boukadi

Abstract: Body Condition Score (BCS) is a key metric for monitoring the health, productivity, and welfare of livestock, playing a crucial role in supporting farmers and experts in effective herd management. Despite advancements in BCS prediction for cows and goats, no computer vision-based methods exist for sheep due to their complex body features. This absence, coupled with the lack of interpretability in existing AI models, hinders real-world adoption in sheep farming. To address this, we propose the first interpretable AI framework for sheep BCS prediction leveraging ontology-based knowledge representation. In this paper, we enrich the ontology MoonCAB, which models livestock behavior in pasture systems, with BCS-related knowledge to prepare it for future integration into explainable AI (XAI) systems. Our methodology involves enhancing the “Herd” module of MoonCAB with domain-specific concepts and 200 SWRL rules to support logical inference. The enriched ontology is evaluated using Pellet, SPARQL, and the MoOnEV tool. As a result, MoonCAB now enables reasoning-based support for BCS-related decision-making in precision sheep farming, laying the groundwork for future developments in ontology-based explainable AI (O-XAI).
Download

Paper Nr: 95
Title:

Semi-Automatic Domain Ontology Construction: LLMs, Modularization, and Cognitive Representation

Authors:

Silvia Lucia Borowicc and Solange Nice Alves-Souza

Abstract: Domain ontology construction is a complex and resource-intensive task, traditionally relying on extensive manual effort from ontology engineers and domain experts. While Large Language Models (LLMs) show promise for automating parts of this process, studies indicate they often struggle with capturing domain-specific nuances, maintaining ontological consistency, and identifying subtle relationships, frequently requiring significant human curation. This paper presents a semi-automatic method for domain ontology construction that combines the capabilities of LLMs with established ontology engineering practices, modularization, and cognitive representation. We developed a pipeline incorporating semantic retrieval from heterogeneous document collections, and prompt-guided LLM generation. Two distinct scenarios were evaluated to assess the influence of prior structured knowledge: one using only retrieved document content as input, and another incorporating expert-defined structured seed terms alongside document content. The approach was applied to the domain of Dengue surveillance and control, and the resulting ontologies were evaluated based on structural metrics and logical consistency. Results showed that the scenario incorporating expert-defined seed terms yielded ontologies with greater conceptual coverage, deeper hierarchies and improved cognitive representation compared to the scenario without prior structured knowledge. We also observed significant performance variations between different LLM models regarding their ability to capture semantic details and structure complex domains. This work demonstrates the viability and benefits of a hybrid approach for ontology construction, highlighting the crucial role of combining LLMs with human expertise for more efficient, consistent, and cognitively aligned ontology engineering. The findings support an iterative and incremental ontology development process and suggest LLMs are valuable assistants when guided by domain-specific inputs and integrated into a structured methodology.
Download

Paper Nr: 176
Title:

Leveraging Large Language Models for Semantic Evaluation of RDF Triples

Authors:

André Gomes Regino, Fernando Rezende Zagatti, Rodrigo Bonacin, Victor Jesus Sotelo Chico, Victor Hochgreb and Julio Cesar dos Reis

Abstract: Knowledge Graphs (KGs) depend on accurate RDF triples, making the quality assurance of these triples a significant challenge. Large Language Models (LLMs) can serve as graders for RDF data, providing scalable alternatives to human validation. This study evaluates the feasibility of utilizing LLMs to assess the quality of RDF triples derived from natural language sentences in the e-commerce sector. We analyze 12 LLM configurations by comparing their Likert-scale ratings of triple quality with human evaluations, focusing on both complete triples and their individual components (subject, predicate, object). We employ statistical correlation measures (Spearman and Kendall Tau) to quantify the alignment between LLM and expert assessments. Our study examines whether justifications generated by LLMs can indicate higher-quality grading. Our findings reveal that some LLMs demonstrate moderate agreement with human annotators and none achieve full alignment. This study presents a replicable evaluation framework and emphasizes the current limitations and potential of LLMs as semantic validators. These results support efforts to incorporate LLM-based validation into KG construction processes and suggest avenues for prompt engineering and hybrid human-AI validation systems.
Download

Paper Nr: 188
Title:

Evaluating Process Parameter Interdependencies Based on Knowledge Graphs in Manufacturing

Authors:

Tom Jeleniewski, Aljosha Köcher, Hamied Nabizada, Jonathan Reif, Felix Gehlhoff and Alexander Fay

Abstract: Formal representations of parameter interdependencies are critical for enabling model-based analysis and reasoning in manufacturing process knowledge graphs. While ontologies based on industrial standards allow for structured semantic descriptions, the computability of embedded mathematical expressions remains a challenging task. This paper presents a SPARQL-driven evaluation framework capable of interpreting and resolving process parameter interdependencies within a knowledge graph. The approach supports an evaluation of nested mathematical expressions, contextual data resolution and computation of process relevant results. The implementation demonstrates how semantic process models can be used for decision support and process optimization tasks.
Download

Short Papers
Paper Nr: 58
Title:

Semantic Rewriting of SPARQL Queries: The Key Role of Subsumption in Complex Ontology Alignments

Authors:

Anicet Lepetit Ondo and Laurence Capus

Abstract: This article introduces an innovative method for rewriting SPARQL queries in the context of complex ontology alignment by leveraging hierarchical relations such as subClassOf and subPropertyOf. The method relies on generalization and specialization links between concepts to retrieve relevant results, even when strict equivalences are missing. In addition, the use of natural language, assisted by the GPT-4 model, helps address the syntactic complexity of SPARQL and facilitates interaction with ontologies. Unlike existing approaches that focus mainly on simple (s: s) or semi-complex (s: c) alignments based on equivalence between source and target entities, our method reinforces semantic matching by explicitly incorporating subsumption relations. It also integrates complex (c: c) correspondences, which are often overlooked in the literature, thereby improving both query coverage and accuracy. Experiments conducted on ontology datasets in the conference domain confirm the method’s ability to capture a wide range of hierarchical relations. While the method is designed to be generic, further evaluations on large-scale ontologies are required to assess its robustness and generalizability.
Download

Paper Nr: 86
Title:

Rule-Based Selection of Languages for Modeling Cyber-Physical Systems

Authors:

Veronica Opranescu and Anca Daniela Ionita

Abstract: In the context of Cyber-Physical Systems (CPS), the choice of suitable modeling languages plays an important role in effectively addressing the varied interests of the system stakeholders. This paper proposes a rule-based recommendation system to suggest appropriate modeling languages that optimally cover all identified viewpoints required for a set of stakeholders. The recommendation engine employs Drools as business rule management system, to highlight the connection between stakeholders’ viewpoints and the kind of models supported by available modeling languages. For assessing this method, a case study was performed with a realistic example from the domain of industrial automation and a selection from three modeling languages.
Download

Paper Nr: 122
Title:

Integrating Retrieval-Augmented Generation with the BioPortal Annotator for Biological Sample Annotation

Authors:

Andrea Riquelme-García, Juan Mulero-Hernández and Jesualdo Tomás Fernández-Breis

Abstract: Integrating biological data remains a significant challenge due to heterogeneous sources, inconsistent formats, and the evolving landscape of biomedical ontologies. Standardized annotation of biological entities with ontology terms is crucial for interoperability and machine-readability in line with FAIR principles. This study compares three approaches for automatic ontology-based annotation of biomedical labels: a base GPT-4o-mini model, a fine-tuned variant of the same model, and a Retrieval-Augmented Generation (RAG) approach. The aim is to assess whether RAG can serve as a cost-effective alternative to fine-tuning for semantic annotation tasks. The evaluation focuses on annotating cell lines, cell types, and anatomical structures using four ontologies: CLO, CL, BTO, and UBERON. The performance was measured using precision, recall, F1-score, and error analysis. The results indicate that RAG performs best when label phrasing aligns closely with external sources, achieving high precision particularly with CLO (cell lines) and UBERON/BTO (anatomical structures). The fine-tuned model performs better in cases requiring semantic inference, notably for CL and UBERON, but struggles with lexically diverse inputs. The base model consistently underperforms. These findings suggest that RAG is a promising and cost-effective alternative to fine-tuning. Future work will investigate ontology-aware retrieval using embeddings.
Download

Paper Nr: 124
Title:

Using AI Models to Generate Knowledge Bases in the Context of IT Service Management

Authors:

Rim Messaoudi, Alexandru Alexandrescu and Francois Azelart

Abstract: Artificial intelligence (AI) models have fundamentally improved its capacity to treat several topics and domains. The purpose of this paper is to use these methods especially Large Language Models (LLMs) to generate knowledge databases in the context of Information Technology Service Management (ITSM). The work starts by extracting information from documents and perform analysis on these documents using Natural Language Processing (NLP) techniques to group them into rigorous knowledge base articles that are easy to classify and search. This paper presents a method that applies generative AI for building a knowledge database in the context of ITSM (Information Technology Service and Management).
Download

Paper Nr: 141
Title:

Foundational Limits: Why BFO’s Aristotelian Framework Cannot Model Modern Science

Authors:

Michael DeBellis

Abstract: The Basic Formal Ontology (BFO) has gained widespread adoption in the biomedical domain and is increasingly promoted as a domain-neutral upper ontology suitable for all branches of science, engineering, and business. Its design reflects a commitment to metaphysical realism rooted in Aristotelian distinctions, particularly between Continuants (entities that persist through time) and Occurrents (entities that unfold over time). While this approach has demonstrated utility in domains such as biology and medicine, it encounters significant limitations when applied to more complex or foundational areas of science, such as quantum physics. Using the example of the electron, whose ontological status defies classical categorization, I argue that the BFO framework lacks the flexibility to accommodate the indeterminacy, contextuality, and non-locality inherent in quantum theory. Bell’s theorem and the incompatibility between general relativity and quantum mechanics further highlight the fragmented and model-dependent nature of contemporary science. These challenges suggest that the search for a single upper model for all domains is based on a mistaken assumption: that science shares a single unified ontology. I conclude that ontology design must acknowledge the methodological and conceptual pluralism of science, and that attempts to enforce a single top-level ontology risk obscuring rather than clarifying the structure of scientific knowledge.
Download

Paper Nr: 142
Title:

Inferring Semantic Schemas on Tabular Data Using Functional Probabilities

Authors:

Ginés Almagro-Hernández and Jesualdo Tomás Fernández-Breis

Abstract: In the information age, tabular data often lacks explicit semantic metadata, challenging the inference of its underlying schema. This is a particular challenge when there is no prior information. Existing methodologies often assume perfect data or require supervised training, which limits their applicability in real-world scenarios. The relational database model utilizes functional dependencies (FDs) to support normalization tasks. However, the direct application of strict FDs to real-world data is problematic due to inconsistencies, errors, or missing values. Previous proposals, such as fuzzy functional dependencies (FFDs), have shown weaknesses, including a lack of clear semantics and ambiguous benefits for database design. This article proposes the concept of functional probability (FP), a novel approach for quantifying the probability of existence of a functional dependency between incomplete and uncertain data, for supporting semantic schema inferencing. FP measures the probability that a randomly selected tuple satisfies the functional dependency with respect to the most frequent association observed. Based on Codd’s relational model and Armstrong’s axioms, this methodology allows for inferring a minimal and non-redundant set of FDs, filtering weak candidates using probability thresholds. The method has been evaluated on two tabular datasets, yielding expected results that demonstrate its applicability. This approach overcomes the limitations of strict dependencies, which are binary, and FFDs, which lack clear semantics, offering a robust analysis of data quality and the inference of more realistic and fault-tolerant database structures.
Download

Paper Nr: 146
Title:

Question Answering over Linked Data with Vague Temporal Adverbials

Authors:

David Maria Schmidt, Svenja Kenneweg, Julian Eggert, Jörg Deigmöller and Philipp Cimiano

Abstract: Vague temporal adverbials are common in human communication but most question answering over linked data (QALD) approaches only work with exact time points. We present a QALD system that interprets vague temporal adverbials (e.g., “just”, “recently”) using a factorized probabilistic model. Building on NeoDUDES, an existing QALD approach, we map vague temporal adverbials to time intervals via empirically grounded Gaussian functions and generate SPARQL queries with temporal filters, enabling compositional interpretation of questions involving vagueness. Evaluated on a knowledge graph based on real-world smart home data, our system shows strong performance.
Download

Paper Nr: 170
Title:

Ontology Semantic Disambiguation by LLM

Authors:

Anastasiia Riabova, Rémy Kessler and Nicolas Béchet

Abstract: Within the BPP project, a combination of statistics and word n-gram extraction enabled the creation of a bilingual (French/English) ontology in the field of e-recruitment. The produced dataset was of good quality, but it still contained errors. In this paper, we present an approach that explores the use of large language models (LLMs) to automate the validation and enrichment of ontologies and knowledge graphs. Starting with a naive prompt and using small language models (SLMs), we tested various approaches, including zero-shot, few-shot, chain-of-thought (CoT) reasoning, and self-consistency (SC) decoding. The preliminary results are encouraging, demonstrating the ability of LLMs to make complex distinctions and to evaluate the relationships derived from our ontology finely.
Download

Paper Nr: 73
Title:

Possibilistic Extension of Domain Information System (DIS) Framework

Authors:

Deemah Alomair and Ridha Khedri

Abstract: Uncertainty poses a significant challenge in ontology-based systems, manifesting in forms such as incomplete information, imprecision, vagueness, ambiguity, or inconsistency. This paper addresses this challenge by introducing a quantitative possibilistic approach to manage and model incomplete information systematically. Ontologies are modelled using the Domain Information System (DIS) framework, which is designed to handle Cartesian data structured as sets of tuples or lists, enabling the construction of ontologies grounded in the dataset under consideration. Possibility theory is employed to extend the DIS framework, enhancing its ability to represent and reason with incomplete information. The proposed extension captures uncertainty associated with instances, attributes, relationships, and concepts. Furthermore, we propose a reasoning mechanism within DIS that leverages necessity-based possibilistic logic to draw inferences under uncertainty. The proposed approach is characterized by its simplicity. It improves the expressiveness of DIS-based systems, introducing a foundation for flexible and robust decision-making in the presence of incomplete information.
Download

Paper Nr: 137
Title:

Enhancing Data Quality and Semantic Annotation by Combining Medical Ontology and Machine Learning Techniques

Authors:

Zina Nakhla and Manel Sliti

Abstract: Effective management of electronic health records (EHR) is a major challenge in the modern healthcare sector. Despite technological advances, the interoperability of medical data remains a crucial challenge. This complex problem is manifested by the diversity of data formats, the presence of multiple standards and the heterogeneity of Information Technology (IT) systems used in health- care establishments. However, the diversity of IT systems and the complexity of medical terminologies often make data interoperability and semantic annotation in the healthcare domain difficult. To address this challenge, our study proposes an innovative approach to standardize the representation of medical concepts, to automate the detection of medical abbreviations and to improve the contextual understanding of medical terms. We developed an ontological model to harmonize the representation of medical data, thus facilitating their exchange and integration between different health systems. In parallel, we used advanced machine learning techniques for automatic detection of medical abbreviations in medical texts, and applied Natural Language Processing to improve contextual understanding of medical terms. The results of our study demonstrate the effectiveness of our approach in solving challenges related to medical data management. By combining different advanced techniques, our approach helps overcome barriers to medical data interoperability and paves the way for better healthcare system integration and improved patient care.
Download

Paper Nr: 178
Title:

Triples-Driven Ontology Construction with LLMs for Urban Planning Compliance

Authors:

Rania Bennetayeb, Giuseppe Berio, Nicolas Bechet and Albert Murienne

Abstract: Ensuring compliance with urban planning regulations requires both semantic precision and fully interpretable decision processes. In this paper, we present a semi-automated methodology that combines the flexibility of large language models with the rigour of Semantic Web technologies to develop an urban planning ontology from regulatory texts. First, the paper presents a systematic evaluation of eight state-of-the-art large language models on the WebNLG dataset for semantic triple extraction task, using few-shot and chain-of-thought prompting. It then discusses the engineering of a domain-adapted prompt. The resulting triples are partially validated through a two-step procedure that takes into account the topological properties of an underlying graph (corresponding to a raw version of a knowledge graph) and the assessment of Human domain experts.
Download

Paper Nr: 180
Title:

OWL-S Grounding Parameters Matching by Means of LLM: Preliminary Investigation

Authors:

Domenico Redavid, Eleonora Bernasconi and Stefano Ferilli

Abstract: SOA architecture was created to systematise issues relating to the interoperability of M2M services, focusing on issues such as security and privacy. With the advent of generative AI, there is a different way to perform the operations for which Semantic Web Services were created, in a much simpler way, but losing control over the level of security and privacy. In this paper, we seek to propose a combined vision of the two approaches, identifying how generative AI can be used to solve specific, rather than general, problems. To this end, we attempt to analyse how an LLM could be used by a software agent to align different types of XML parameter data in WSDL descriptions.
Download