Purpose: Existing relaxation works either focus on ontology-based or entity-based elements, leaving a gap in handling queries that combine both. The purpose of this paper is to fill this gap and relax real-world SPARQL queries combining ontology-based and entity-based techniques to retrieve top-k relevant results. Methodology: We propose a hybrid query relaxation solution with a ranking system based on information content, integrating ontology-based and entity-based strategies by extending similarity measures beyond classes to entities and literals. Findings: The model retrieves relevant top-k results and better aligns with user expectations across diverse query types. It effectively ranks relaxed queries involving ontology-based and instance-based relaxations. Value: This work introduces a unified relaxation framework and an evaluation benchmark grounded in human-judgment-based relevance. It advances the literature by providing a relaxation system that aligns closely with user-validated relevance, especially for queries involving annotations.
Purpose: The purpose of this paper is to determine whether a domain-based bias exists within entity linking techniques and how to potentially mitigate it. Methodology: To study how annotation results of entity linkers are domain-dependent, we annotate 6 widely used benchmarks from varying domains with 12 entity linkers and identify 37 distinct topics. We adopt a predictive approach that leverages topic modelling to assess system performance across domains determine whether domain information by itself is a sufficient signal to determine an optimal system. Findings: By analysing the interaction between domains and annotation behaviour, we find that no single system consistently outperforms across all domains. Instead, performance varies significantly depending on domain, suggesting that optimal results require domain-specific system selection. Value: To detect and mitigate domain effects, we provide our approach, tools, and benchmarks for enhanced domain-balanced (DOMiNO) and domain-aware (DOMeX) evaluation. Our findings underline the critical role of domain awareness in the development and deployment of text-processing systems, providing a pathway for more adaptable and robust methodologies. We release and open source all generated data, code and findings for future research endeavours on our repository at https://github.com/kmdn/linking_bias/.
Purpose: Ontology Matching (OM) has been studied for decades, yet fully automatic solutions remain elusive because ontologies differ in structure, granularity and vocabulary. Nevertheless, the abundant textual content attached to ontology entities suggests that the task could benefit from modern language-representation models. We therefore present the Semantically-Informed Similarity Matching Algorithm (SISMA), a novel system that matches concepts by leveraging the similarity of SBERT embeddings computed over pseudo-sentences extracted from the ontologies. Methodology: We focus on the task of class and property equivalence. We represent each ontology concept as a set of SBERT embeddings associated with each predicate. For every pair, a similarity matrix is computed and reduced to a score via linear operations with two learnable matrices. These are trained on a dedicated dataset. We evaluated our system on the OAEI benchmark alignments, training on the Conference track and testing on the Circular Economy (CE) and Material Sciences and Engineering (MSE) tracks. Findings: Our experiments reveal that the SISMA method achieves performance comparable to the state of the art. On the CE track our system achieves a higher F1-score than the participating systems, while on the MSE track it performs slightly lower. We also compared our results with a baseline across the parameter space, confirming that the training step is key to overall performance. Value: We have designed, implemented, and evaluated a novel system for ontology matching that achieves performance comparable to state-of-the-art methods. Our approach is readily extensible-primarily by training and testing on additional datasets-and the underlying idea can be realized in alternative ways, for example by replacing the current linear-operator scoring and threshold-filtering approach with a classifier that operates directly on the similarity matrix space.
Purpose: Existing approaches for generating Knowledge Graphs (KGs) from file-based, non-RDF data using declarative mappings are either limited by language-specific engines or lack optimization with language-independent relational algebra backends, resulting in suboptimal performance. This research proposes an integrated framework that tightly couples logical and physical plan optimizations, enabling high-performance, language-agnostic RDF graph construction. Methodology: We formalize the KG construction process using relational algebra with a dedicated RDF term generation function within the projection operator, resulting in one of two canonicalized logical plans, one with a join and one without. We then introduce tightly coupled physical operators used to define concrete execution pipelines. We propose and evaluate two optimizations, logical-level constant-folding to reduce redundant computations and a physical-level heuristic scheduling strategy to optimize concurrent execution. We implemented the optimizations in a new backend engine called konverter and benchmarked the engine with an RML frontend against two comparable engines, Morph-KGC and FlexRML. Findings: Empirical results show that constant-folding improves performance by approximately 7.4% and heuristic scheduling by approximately 14.7% compared to a worst case scenario, with minimal additional memory overhead. Overall, konverter outperforms the current state-of-the-art in performance FlexRML, reducing execution time by 61.5% and peak memory usage by 25.1%, though it currently only supports CSV files, a limitation we aim to address in future work. Value: The proposed framework and optimizations provide a formal and practically validated approach to optimizing the execution of declarative mappings for KG construction. The konverter engine demonstrates the potential for building high-performance, language-agnostic engines for enterprise KG construction.
Purpose: The heterogeneity of scholarly information in knowledge graphs (KGs) and unstructured textual sources poses challenges in building robust Scholarly Question Answering (SQA) systems. Existing datasets and models typically address a narrow spectrum, focusing exclusively on KGs or unstructured sources and limiting evaluation to simple factoid questions. This gap leaves current systems unable to answer complex, hybrid scholarly questions that require integrating evidence from multiple heterogeneous data sources. Methodology: We introduce HySQA (Hybrid Scholarly Question Answering), a large-scale benchmarking dataset containing hybrid questions over scholarly KGs and Wikipedia text. HySQA contains complex questions that need to traverse facts across structured and unstructured sources. We also develop a baseline model that adaptively decomposes each question into sub-questions, identifies their answer sources, retrieves relevant information from SKGs and Wikipedia, and generates an answer using a hybrid augmented answer generation framework. Findings: The experimental results show that integrating static and adaptive decomposition methods is more effective than static decomposition alone. Value: Introducing HySQA provides the community with resources for evaluating the advancements in scholarly QA research.
Automatic term extraction (ATE) identifies domain-specific concepts from specialized corpora, but suffers from limited annotated training data across diverse domains. We propose three novel LLM-based data augmentation schemes for ATE: context-level augmentation (generating diverse sentences using existing terms), term-level augmentation (replacing terms with domain-relevant alternatives), and combined augmentation (creating novel sentences with new terminology). Our approach leverages both ChatGPT-4o and Wikipedia-derived domain lexicons to generate synthetic training data. Experiments across four domains in the ACTER dataset demonstrate consistent improvements over state-of-the-art XLM-RoBERTa baselines, with gains of up to 28% F1-score in few-shot scenarios (5-10 samples) and 1-2% improvements in larger datasets (100-500 samples). Context-level and term-level augmentation consistently outperform combined augmentation, while LLM-based methods surpass Wikipedia-based augmentation. Our findings establish the effectiveness of targeted data augmentation for ATE across varying data availability scenarios, with performance gains extending beyond few-shot settings to practical dataset sizes.
We propose the RISKMAN ontology and shapes for representing and analysing information about risk management for medical devices. Risk management is concerned with taking necessary precautions to ensure that a medical device does not cause harms for users or the environment. To date, risk management documentation is submitted to notified bodies (for certification) in the form of semi-structured natural language text. We propose to use terms from the RISKMAN ontology to provide a formal, logical underpinning for risk management documentation, and to use the included SHACL constraints to check whether the provided data is in accordance with the requirements of the two relevant norms, i.e. ISO 14971 and VDE Spec 90025. RISKMAN is available at https://w3id.org/riskman.
Knowledge graphs (KGs) are a key tool across fields like web search, healthcare, and technical assistance but are known to be notoriously incomplete. This paper addresses open-world knowledge graph completion (OW-KGC), which involves linking newly emerging entities - which are mentioned in text but not yet present in the graph - to a KG. We investigate the potential of off-the-shelf LLMs to tackle this task. Here, a central challenge is to constrain the LLM to the candidate space of existing entities available for linking. To address this, we propose a two-stage pipeline: A lightweight weight pre-ranker fine-tuned on the particular task narrows down the candidate entities, which are then re-ranked by the LLM. We evaluate two recent open-weight LLMs, namely Meta-LLAMA3 Instruct (70B) and DeepSeek-R1-Distill-LLAMA (70B), on public OW-KGC benchmarks for the KGs Freebase (FB15k237), WordNet (WN18RR), and Wikidata (IRT2). We test two variants of the pipeline: one where the LLM is applied as a strict re-ranker, and another where the LLM is allowed to provide additional suggestions. Our experimental results show that, while LLMs used in isolation perform poorly, they improve performance substantially when applied as re-rankers, achieving new state-of-the-art results across all datasets. Notably, we observe that DeepSeek, which leverages an explicit self-reflection mechanism before producing results, outperforms LLAMA significantly when re-ranking and measuring MRR. However, LLAMA contrary to DeepSeek, is considerably better at accessing and retrieving correct answers from its internalized knowledge.
Purpose: This paper examines RDF validation practices and challenges to understand stakeholder applications, their needs, and identify areas for improvement in technologies and methodologies, thereby guiding future research and standardization efforts. Methodology: A community survey was conducted, targeting a diverse group of RDF validation technology users across academia and industry. The survey collected data on current practices, tool usage, perceived benefits, limitations, and desired enhancements to gain a broad overview of the validation landscape. Findings: Our analysis shows that while RDF validation is widely adopted and valued for enhancing data quality, significant challenges remain. In particular, users report a need for better documentation, improved tool support, enhanced performance, and greater language expressiveness to handle complex large-scale validation tasks effectively. Value: This work provides crucial insights into the RDF validation landscape, highlighting current practices and key areas for development. It offers a foundation for researchers, developers, and standardization bodies to address current limitations and advance validation technologies, ultimately improving data quality and usability in knowledge graphs.
Purpose: This paper addresses the lack of practical support for generating RDF Collections and Containers from heterogeneous sources in existing mapping tools with the RDF Mapping Language (RML). While the RML Collections and Container (CC) module defines their generation, RML-CC implementations remain limited to a reference implementation called BURP that was not conceived for efficiency or scalability. We aim to close this gap by extending a tool for efficient RML generation with support for RML-CC. Methodology: We extended Morph-KGC to support RML-CC and developed YARRRML-CC for user-friendly mapping definitions. We also updated Yatter to enable translation from YARRRML-CC to RML-CC. We validated these tools using 35 RML-CC test cases and 22 additional YARRRML-CC cases and conducted performance evaluations using synthetic datasets. Findings: While BURP passes all RML-CC test cases and scales up to 1M records, Morph-KGC passes only 51% due to architectural constraints and struggles with larger datasets-Morph-KGC's reliance on Pandas limits support for complex constructs such as nested collections. Yatter fully supports YARRRML-CC. Morph-KGC performs well in standard RDF generation but struggles with RML-CC at scale. This highlights the importance of selecting tools that align with the structural complexity and performance demands of specific use cases. Value: Our work enhances the practical applicability of RML-CC in knowledge graph construction by providing different tooling supports (BURP for the Java ecosystem, and Morph-KGC for complete Python pipelines), interoperability through YARRRML-CC, and validated performance insights.
Purpose: Semantic device descriptions, such as the Web of Things (WoT) Thing Description (TD), are a great tool to improve the level of interoperability in Internet of Things (IoT) systems. However, a majority of new and existing IoT devices do not ship with a TD, and developers often need to create them by hand. This makes it tedious for anyone who wishes to integrate, migrate, or modernize devices of their existing infrastructure into a WoT ecosystem. Therefore, an automated approach for TD generation that facilitates this process is needed. Methodology: We propose a Large Language Model (LLM)-based approach to automate TD generation. By utilizing message logs and conformance checks, we introduce an iterative process that leverages LLM technologies to generate TDs. The proposed methodology is evaluated in a case study of 76 IoT devices communicating over MQTT. Findings: Our results show that with the proposed methodology, an LLM can generate TDs from MQTT message logs with an average functional accuracy of up to 91%, and a descriptive accuracy of around 85%, demonstrating strong overall performance. Value: All generated TDs and the prototypical Python implementation of the methodology can be found in our repository. The proposed methodology helps the adoption of the WoT by offering an automated generation of TDs in environments where MQTT message logs are available.
Purpose: The individualized delivery of educational content benefits from the usage of various multimedia formats. Hence, there is a need for a tool that supports visual engagement, especially in cases of video consumption. Methodology: In this short paper, we present an extension to the Interactive Knowledge (InK) Browser that accommodates the exploration of educational multimedia material. The InK Browser already provides an intuitive and interactive interface that allows a follow-your-nose data discovery of a knowledge graph. Additionally, we provide the InK Browser with a new functionality that supports the visual parsing of video content and links it to openly accessible educational material. Findings/Value: Additionally, we report results from a Survey. Finally, we propose an underlying ontology that reflects this design and encourages both modularity and re-usability in the representation of knowledge.
Purpose: The purpose of this paper is to explore whether ontologies can improve LLM-based document-level relation extraction (docRE) for knowledge graph (KG) construction. In particular, the authors are interested in harnessing semantics in LLM-supported docRE for KG construction and retrieving the docRE-relevant ontology segment of an ontology. Methodology: To explore how ontologies can improve LLM-based docRE for KG construction, the paper provides GrOWL-RAG, a framework consisting of an Ontology RAG to provide structured domain knowledge and reasoning capabilities as prompt context. The authors evaluate GrOWL-RAG on Re-DocRED benchmark data with promising precision results and conduct an ablation study with LLMs. Findings: The paper shows value add of GrOWL-RAG under an ablation study. In this paper, the authors identify multiple fields for further development of this approach, such as improving performance, extending GrOWL-RAG with ontology embeddings, and exploring different ontology types. Value: Integrating ontologies into RAG is a relatively unexplored field in which this paper shines light on for the task of docRE. Further, GrOWL-RAG is an LLM-based approach for docRE without the need for training or finetuning. The authors publish code and input files along with a new ontology for the docRE test set Re-DocRED to support docRE research utilizing ontologies. This paper sets the stage for further discussions of this topic.
We present a method for integrating visual schema diagrams and visual queries within a single interactive environment for knowledge graph exploration. This approach addresses the gap between schema visualization tools and visual query interfaces, thereby enabling users, including domain experts, to seamlessly perform schema-based queries without the need for switching between different tools. We implement the method in the ViziQuer tool, which provides also means for full visual queries alongside a schema-based querying. The option for a transition from schema-based queries to the full visual query environment is provided as well.
Many domains use specific terminologies to describe concepts. Being able to explicitly manage such terminologies instead of relying on their common knowledge is beneficial both for newcomers and for people for whom the terminology is not their daily bread and butter. This is especially true for legislative terminologies. We present TermIt, a Semantic Web-based terminology manager that allows domain experts to create and manage high-quality terminologies, link them to normative documents as well as use them to annotate other related documents. We discuss the architecture of the system and the technologies used in its development.
Cross-border railway travel in Europe continues to face significant challenges due to fragmented digital ecosystems, incompatible data standards, and regulatory inconsistencies, despite progress in national railway markets. This study set out to investigate how data interoperability and emerging data-sharing ecosystems-particularly European Rail Data Spaces-can address these persistent barriers. Through stakeholder focus groups with national operators (OBB, SBB, DB), expert interviews, and a comparative analysis of existing data models such as Transmodel, GTFS, and NeTEx, the research identified key organizational and technical bottlenecks. The findings demonstrate that while standardized data formats have the potential to enhance interoperability, their fragmented adoption and lack of harmonized implementation across countries limit their effectiveness. To address these issues, the study proposes a theoretical framework for seamless cross-border rail integration, emphasizing the critical role of trusted, interoperable data environments in supporting a unified European railway network.
The rapid growth of renewable energy toward industry 4.0, particularly renewable solar, has amplified the need for findable, accessible, interoperable, and reusable data sets collected from meteorological stations, sensors, renewable equipments, infrastructures, and others. In particular, solar plants generate large volumes of operational data, often structured using the IEC 61850 standard, a widely adopted protocol for energy systems that defines models for communication between intelligent devices. While IEC 61850 ensures consistency at the data exchange level, it does not provide full semantic interoperability needed for advanced analytics, data discovery, cross-domain integration, and automated reasoning. This paper describes an automatic pipeline for transforming IEC 61850-tagged data into knowledge graphs, enabling seamless integration with diverse datasets such as weather and grid structure for operation maintenance services as well as energy prediction services. This pipeline allows the transformation of these tags into a KG aligned with the Omega-X ontology, reducing the need for manual intervention, enabling continuous data integration, and enhancing the reuse of these data by service providers in the energy sector. We evaluate this approach in the context of the Omega-X project, using real-world datasets from a solar park that combines meteorological and electrical parameters.
The digital transformation of engineering systems demands scalable and precise identifier management. Information about a single asset is often fragmented across numerous systems and organizational boundaries, with each context using its own identifiers. This paper addresses this challenge by introducing a formal framework for semantic reference. We build on prior work by introducing two key concepts: reference contexts, which formalize the boundary conditions for identifier interpretation, and public models, which serve as curated, shared layers for anchoring reference. We define reference equality as the symmetric, transitive closure over typed proxy relations that link identifiers across these contexts. Finally, we demonstrate how this semantic infrastructure provides a foundational component for a new generation of industrial systems, including providing essential grounding for Large Language Model (LLM) workflows. This approach bridges the gap between human-readable descriptors and machine-readable identifiers, aligns with modern architectural principles like Data Mesh, and supports hybrid reasoning in industrial knowledge systems.
This paper describes prototyping experiences in a population health use case of the draft Health DCATAP specification for health data catalogues under the European Health Data Spaces Regulation (EHDS). Using data catalogues to support data-driven health planning like this is an important use case. Our work included the development of a data catalogue metadata model, catalogue record creation via direct data entry and scraping of open data, and development of record quality and feasibility reports. It was found necessary to extend the catalogue with new classes and properties for this use case, some of which were from the Data Privacy Vocabulary (DPV), and a number of limitations in the current Health DCAT-AP specification draft were discovered. Stakeholders were generally positive in their assessment of the contribution of this novel structured approach to health data indicator discovery and assessment. This shows the potential for the semantic data governance infrastructure specified by the European Health Data Spaces Regulation to influence future data-driven decision making at all levels of European health services. The catalogue metadata model, report queries and data scraping code are all made available as open source resources for reuse by others. One new property has been added to DPV as a result of this work and it will feed into the Health DCAT-AP standardisation process in the ETSI/TC Data. This paper describes a population health use case based on defining a health and wellbeing profile for older adults, data catalogue competency questions for this use case, a metadata model for the catalogue that meets these requirements, and a data quality feasibility and assessment reporting workflow along with stakeholder feedback.
