This article investigates the methodological foundations and theoretical assumptions behind the construction of computational ontologies for modeling narrative and fiction, with a focus on literary characters. We survey and critically assess a set of existing domain-specific ontologies for fictional narrative, evaluating their modeling strategies, taking into consideration their philosophical and knowledge representation criteria. Drawing from ontology engineering principles and foundational frameworks such as Descriptive Ontology for Linguistic and Cognitive Engineering (DOLCE) and Basic Formal Ontology (BFO), we propose a two-class ontology mapping methodology (harmonisation and alignment) to evaluate and foster semantic interoperability across the considered models. An experimental ontology pattern for fictional characters is then introduced and aligned with both DOLCE and BFO via International Committee for Documentation Conceptual Reference Model (CIDOC-CRM), revealing the ontological commitments and modelling trade-offs required to formalise the nuanced nature of fictional entities. This study offers a preliminary attempt to explore how foundational ontologies might support conceptual clarity, while also highlighting the epistemological challenges involved in representing complex, non-referential cultural artefacts. Ultimately, this work aims to highlight the relevance of ontologies as a shared infrastructure for computational literary studies, supporting interdisciplinary collaboration, fostering Open Science and encouraging more structured, transparent, and conceptually grounded approaches to the representation and analysis of cultural phenomena.
To achieve interoperability for effective data exchange on the web, we need a contract. Depending on the field, we may refer to technical data schemas, web vocabularies, or, more generally, to data specifications (DSes). The development and management of these DSes can become difficult in complex domains with multiple stakeholders and related DSes involved. In this article, we present Dataspecer, an open-source, modular web application for the development of semantic data specifications (SDSes), DSes that target the semantic and technical layers of data exchange. Dataspecer allows users to design web vocabularies and their application profiles, maintaining relations between reused concepts and their original SDSes. Furthermore, Dataspecer assists users in the creation of technical artifacts such as schemas for JSON or XML, while maintaining consistency of the artifacts with the application profiles. We motivate the need for SDSes and derive requirements for such a tool. In case studies based on the ecosystem of Data Catalog Vocabulary-based specifications, we demonstrate that SDSes created in Dataspecer meet these requirements and are of higher quality. We show SDSes that were created directly in Dataspecer, and in the evaluation section, we argue that using our tool is more efficient than creating them manually, even for smaller domains.
Large language models (LLMs) have shown effectiveness in various natural language understanding (NLU) tasks. However, they face notable limitations like hallucinations, a lack of contextual knowledge, and outdated or incomplete knowledge when applied across knowledge-intensive domains such as scientific research, biomedical sciences, finance, law, and others. These challenges commonly arise from the scarcity and under-representation of domain-specific data during the training and model alignment phases. Furthermore, Large Language Models (LLMs) struggle to provide nuanced expertise, as their internal knowledge remains static and generalized, hindering their ability to reason accurately or deliver context-aware results in specialized tasks. This survey investigates the integration of external knowledge into LLMs to address these limitations. The focus is on decoder-based LLMs, that is, autoregressive models that generate text sequentially. By investigating parametric and non-parametric approaches, this work discusses methods to enhance model reasoning capabilities, factual accuracy, and adaptability for domain-specific and knowledge-intensive tasks. Additionally, it highlights the potential of integrating external knowledge to improve explainability and ensure more trustworthy outputs. This survey supports software developers and natural language processing (NLP) researchers in designing NLU systems for specialized domains by leveraging pre-trained LLMs. Additionally, the work provides a foundation for advancing LLM-based NLU systems with insights into future research areas.
This article introduces CAAPT (Computational Approaches to Addressing Problematic Terminology), an ontological framework for the representation of cultural heritage terminology guidance documents and the decision-making practices involved in this domain as linked open data. CAAPT consists of three constituent ontologies: CAAPT-O, CAAPT-UC, and CAAPT-DM. These are trialed through the instantiation of a knowledge graph populated by the contents of three cultural heritage terminology guidance documents, three sources documenting critical cataloging decisions and actions, and a related linked open data project vocabulary. This knowledge graph demonstrates the affordances of the ontologies. A linked open data vocabulary, CAAPT-V, is also introduced in order to provide a set of reference values to be used in instantiations of the ontological framework. Lastly, this article proposes a novel approach to ontology engineering that is grounded in critical theory, namely concepts from feminist and queer theories, thus aligning the theoretical framework of the technical development work with that of the domain it is considering.
We present the extended NeOn-GPT pipeline, an LLM-powered, domain-agnostic ontology learning framework grounded in the NeOn methodology. The pipeline comprises two components: (i) ontology draft generation through multi-step prompting-covering requirement specification, competency questions, conceptualization, formal modeling, population, and documentation-and (ii) automated verification and repair through orchestrated calls to third-party tools complemented by LLM-suggested fixes. The extended pipeline introduces an explicit ontology reuse step to guide LLMs toward more consistent modeling decisions. We evaluate NeOn-GPT across four domains (Wine, Cheminformatics, Environmental Microbiology, and Sewer Networks) using both proprietary (GPT-4o) and open-source (Mistral, Llama-4, DeepSeek) models. Gold-standard alignment is assessed via structural metrics (class, property, and axiom profiles), lexical metrics, and semantic metrics based on sentence embeddings. Results show that LLMs consistently generate ontologies with rich relational structures and meaningful semantic alignment, with most entity and triple similarities falling in the 0.5-0.8 range. This study provides a comprehensive, cross-domain evaluation of NeOn-guided LLM ontology learning, clarifying its capabilities and limitations.
The choice made for representing the inputs and outputs of generative pretrained language models (PLMs) can impact their fine-tuning on a new task. This article focuses on the fine-tuning and linearization process to generate facts extracted from text. On a restricted relation extraction (RE) task, we challenged five encoder-decoder models, including BART, T5, CodeT5, FlanT5, and PileT5 by fine-tuning them on 13 linearization variations, including RDF standard syntaxes and variations thereof. Our benchmark covers the validity of the produced triples, the model's performance, the training behavior, and the resources needed. We show these PLMs can learn some syntaxes more easily than others, and we identify a promising "Turtle Light" syntax supporting the quick and robust learning of the RE task.
The verbalization of structured data is a beneficial process for several applications. In the context of knowledge graphs (KGs), transforming Resource Description Framework (RDF) triples into natural language facilitates tasks such as KG documentation or alternative exploration methods for different user needs. While significant progress has been made on the English verbalization of KGs, Spanish remains an under-represented language for this task due to the lack of suitable resources. This hinders the development and evaluation of models capable of generating high-quality Spanish verbalizations. To tackle this problem, we create a Spanish adaptation of the WebNLG dataset, a benchmark consisting of over 45,000 verbalizations paired with DBpedia triple sets. To our knowledge, this is the first formal attempt to provide such a dataset in Spanish, which not only serves for data verbalization but can also potentially support the automated generation of RDF triples from text. We leverage this dataset to conduct a comprehensive evaluation of resource-efficient models for the Spanish triple-to-text task employing two different learning approaches: context learning (zero-shot, one-shot, and few-shot settings) and supervised learning through partial fine-tuning. Our results highlight the challenges of generating fluent and accurate Spanish text and demonstrate that partial fine-tuning of the evaluated models significantly improves performance.
Scientific processes are often described in free text, making it difficult to represent and reason over them computationally. We present schema-miner p r o , a human-in-the-loop framework that automatically extracts and grounds structured schemas from scientific literature. Our approach combines large language models for schema extraction with an agent-based system that aligns extracted elements to external ontologies through interpretable, multi-step reasoning. The agent leverages lexical heuristics, semantic similarity, and expert feedback to ensure accurate grounding. We demonstrate the framework on two semiconductor manufacturing workflows—atomic layer deposition and atomic layer etching—mapping process parameters and outputs to the QUDT (Quantities, Units, Dimensions, and Types) ontology. By producing ontology-aligned, semantically precise schemas, schema-miner p r o lays the groundwork for machine-actionable scientific knowledge and automated reasoning across disciplines.
Most existing arithmetic word problem (AWP) solvers focus on solving simple examples. Transfer case-AWPs (TC-AWPs) involve scenarios where objects are transferred between agents. The widely used AWP datasets mainly consist of simple TC-AWPs (problems that involve a single object transfer). Current large language models (LLMs) are capable of solving most of these simple TC-AWPs effectively. In this work, we focus on assessing the solving capability of LLMs (ChatGPT and Gemini) for complex TC-AWPs (where multiple types of objects are transferred or more than one transfer of an object is performed). Since the popular AWP datasets contain only simple TC-AWPs, we first generate complex TC-AWPs using an ontological approach. We utilize these complex examples to assess LLMs' word-problem-solving capabilities. We observe that the accuracy of LLMs falls down rapidly as the number of object transfers is increased to 3 or 4. An approach for solving TC-AWPs using ontologies and M/L exists in the literature. We propose an extension of this approach that can handle complex TC-AWPs and find that, compared to the current LLMs, the proposed solution gives better accuracy for complex TC-AWPs. We analyze the failed cases of the LLM approach and find that the reasoning capabilities of LLMs need a lot of improvement.
There are a few applications available for educational purposes in the forestry domain. These applications have significant limitations, including not exploiting existing biodiversity datasets, lacking flexible and consistent use of domain concepts, and generating annotations that are not easily shareable or reusable by other applications. In this paper, we introduce EducaWood, a novel Semantic Web application designed for forestry education that overcomes these limitations by leveraging Linked Open Data (LOD). Users can easily create tree annotations through a web form that hides the complexity of Semantic Web technologies. These annotations adhere to the Simple Tree Annotation ontology and are saved in a triplestore, facilitating seamless sharing with other users and applications. Moreover, EducaWood offers scalable and efficient visualization of semantic tree data across various zoom levels on a map interface. Access to LOD is handled through a REST API that allows read and writes operations over multiple data sources. An implementation of EducaWood has been successfully tested by more than 650 users, including real students and teachers in a pilot educational experience.
The paper analyzes and characterizes the algebraic and logical structure of the multiset semantics for SPARQL patterns involving AND, UNION, FILTER, EXCEPT, and SELECT. To do this, we align SPARQL with two well-established query languages: Datalog and Relational Algebra. Specifically, we study (i) a version of non-recursive Datalog with safe negation extended to support multisets, and (ii) a multiset relational algebra comprising projection, selection, natural join, arithmetic union, and except. We prove that these three formalisms are expressively equivalent under multiset semantics.
This paper presents a Shapes Constraint Language (SHACL)-based framework for validating the Time Ontology (https://www.w3.org/TR/owl-time). The Time Ontology, currently a W3C Candidate Recommendation, is widely recognized as the 'de facto' standard for representing temporal data in the Semantic Web. However, its current OWL axiomatization cannot enforce several validation constraints on temporal knowledge that can be expressed using the Time Ontology vocabulary. These constraints are instead captured by the SHACL formalization proposed in this paper. Nevertheless, we show that SHACL shapes are insufficient to validate even simple knowledge graphs that can be encoded using this vocabulary. This limitation arises because validation must be performed on the inferred knowledge graph, which SHACL shapes alone cannot derive internally. To address this, our framework first computes the inferred knowledge graph using SHACL-SPARQL rules and then validates it through SHACL shapes. We argue that our findings extend beyond the Time Ontology and have broader implications for SHACL and knowledge graph reasoning. We therefore view our work as a call to action for the Semantic Web community to systematically investigate the interplay between validation and inference. Specifically, there is a need to study the representational requirements of different use cases to identify the minimal set of SHACL shapes and inference rules for data validation in each context. These efforts could ultimately lead to the definition of distinct SHACL dialects, analogous to how OWL Lite, OWL DL, and other profiles were defined for OWL. The shapes and rules that define the proposed framework are available at https://github.com/liviorobaldo/TimeOntologyInSHACL.
With the increase in the food allergic population worldwide, food allergen traceability has become an imperative food safety concern. Food businesses, however, have difficulty ensuring food allergen traceability because it is time-consuming and costly to obtain accurate food allergen data along the supply chain. Semantic Web technologies have great potential to improve efficiency and accuracy of food allergen traceability through automating food data exchange along the supply chain. In this paper, we present the Food Allergen Traceability Ontology (FATO), the first ontology that focuses on food allergen management and traceability processes. To overcome the overspecification problem in the development of ontologies, we propose the integration of a range of knowledge sources on improving food allergen management, in addition to domain experts, to inform the development of FATO. The ontology builds on and is compatible with existing food and product ontologies and models, and captures knowledge on food allergen declarations, food allergen management processes, and traceability. Application examples are provided to illustrate how FATO can be employed to address long-standing issues in food allergen management as well as drive innovation in food businesses.
Unhealthy diets are a major modifiable risk factor for noncommunicable diseases and the leading cause of morbidity and mortality world-wide. Nutrition assessment is crucial for understanding and preventing unhealthy diets, but often relies on self-reporting, which is burdensome, error-prone, and ineffective for long-term tracking. Digital receipts from loyalty cards, enriched with product nutrition information, offer a promising alternative. Current regulations support user-consented access to such receipts and mandates food nutrition information provision, creating a viable legislative foundation for sharing and using digital receipts. Shared ontologies enable consistent management and exchange of digital receipts and food product information across sources and applications. Yet, no current ontology describes enriched digital receipts at product and basket levels with detailed nutrition metrics. We present NutriLink, an ontology connecting digital receipts to product nutrition information and structured dietary recommendations. NutriLink supports evaluating nutritional quality of purchases within and across baskets and delivering structured dietary recommendations. It integrates with the established ontologies, including FoodOn, GoodRelations, and AGROVOC, as well as with schema.org concepts. We demonstrate NutriLink's value through deployment in a fully automated dietary counseling system with 76 users. NutriLink is freely and openly available for research and practice.
The proliferation of large and ever-growing resource description framework (RDF) datasets has sparked a need for robust and performant RDF archiving systems. In order to tackle this challenge, several solutions have been proposed throughout the years, including archiving systems based on independent copies, time-based indexes, and change-based approaches. In recent years, modern solutions combine several of the above mentioned paradigms. In particular, aggregated changesets of time-annotated triples have showcased a noteworthy ability to handle and query relatively large RDF archives. However, such approaches still suffer from scalability issues, notably at ingestion time. This makes the use of these solutions prohibitive for large revision histories. Furthermore, applications for such systems remain often constrained by their limited querying abilities, where SPARQL is often left out in favor of single triple-pattern queries. In this article, we propose a hybrid storage approach based on aggregated changesets, snapshots, and multiple delta chains that additionally provides full querying SPARQL on RDF archives. This is done by interfacing our system with a modified SPARQL query engine. We evaluate our system with different snapshot creation strategies on the BEAR benchmark for RDF archives and showcase improvements of up to one order of magnitude in ingestion speed compared to state-of-the-art approaches, while keeping competitive querying performance. Furthermore, we demonstrate our SPARQL query processing capabilities on the BEAR-C variant of BEAR. This is, to the best of our knowledge, the first openly available endeavor that provides full SPARQL querying on RDF archives.
The use of Semantic Web in cybersecurity systems is becoming more and more popular. This is an important problem, especially in times when Internet of Things (IoT) systems are developing very quickly and and their security must be maintained. Thanks to the semantic web, it is possible to store and process cybersecurity knowledge using ontology. We describe a system for analyzing the level of cybersecurity among Polish citizens, in particular IoT users. An ontology-based knowledge representation related to the security level was created for the described system. The ontology contains the information necessary to determine the security level in different locations and to conduct deeper analysis. It has been prepared for the needs of the IoT system for storing data and knowledge. The described Semantic Web application is part of a larger project that allows to determine cyber security and cyber threats of IoT devices.
Sharing datasets that change (through creates, updates, deletes) poses challenges to data consumers, including reconciling historical versioning and managing frequent changes. This is evident for Knowledge Graphs (KGs), materialized from such datasets, where synchronization happens through frequent regeneration. However, this is time-consuming, loses history, and wastes computing resources through redundant processing. We present a KG generation approach that efficiently handles evolving data sources with different change signaling strategies. We investigate change signaling strategies of real-world datasets, propose corresponding change detection algorithms, and introduce a declarative approach based on the RDF Mapping Language (RML) and Function Ontology to materialize changes for evolving KGs. Detected changes can be automatically published as a Linked Data Event Stream (LDES), using the Activity Streams 2.0 vocabulary to describe changes and communicate them over the Web. We implement our approach in the RMLMapper as Incremental RML and evaluate it both functionally, and quantitatively using a modified version of the GTFS Madrid Benchmark and several real-world data sources. Our approach reduces storage and computing requirements for generating and storing multiple KG versions (up to 315.83x less storage, 4.59x less CPU time, and 1.51x less memory) and reduces KG construction time up to 4.41x. Performance gains are more pronounced for larger datasets, while our approach's overhead partially offsets benefits for smaller ones. Overall, our approach lowers the cost of publishing and maintaining KGs and, via LDES, supports timely, Web-native dissemination of changes. We plan to optimize our change detection algorithms and use windowing to support streaming data.
Knowledge graphs from text have garnered substantial interest across various domains due to their potential to facilitate efficient information retrieval and knowledge exploration. However, knowledge graph generation from textual sources presents unique challenges, particularly in the Islamic domain, where primary sources of knowledge are texts in Arabic, which exhibit complex linguistic and cultural nuances. This article presents a comprehensive methodology for generating a knowledge graph from the hadith corpus. Hadith, a fundamental resource in the Islamic domain, stands as one of the primary sources of Islamic legislation, encompassing the sayings, actions, and silent approvals of the Prophet Muhammad. Leveraging Natural Language Processing techniques, we systematically extract, annotate, and interlink semantic entities and relationships from the hadith corpus, extend the SemanticHadith ontology for entity organization, and compute textual similarities to establish semantic connections. We generate a comprehensive knowledge graph by applying these methods to six hadith collections, facilitating efficient information retrieval and knowledge exploration in the Islamic domain. This is an essential step towards annotating and linking the hadith corpus to allow semantic search to support scholars or students in creating, evolving, and consulting a digital representation of Islamic knowledge. The SemanticHadith knowledge graph is freely accessible at http://www.semantichadith.iknex.com/.
This article introduces SPARQLuedo and OWLuedo, two open-source educational resources designed for hands-on learning of Semantic Web technologies: SPARQL and OWL. Inspired by the board game Cluedo, these resources challenge learners to act as investigators solving a murder. SPARQLuedo guides users in formulating SPARQL queries to interrogate a dedicated RDF knowledge graph and uncover details of the crime, including the victim, the murderer, the location, and the murder weapon. OWLuedo, on the other hand, prompts learners to extend an existing ontology to model the crime scene in greater depth and leverage an OWL reasoner to identify the culprit. These resources, intended to complement lectures, aim to make learning Semantic Web technologies more engaging and interactive. Positive feedback from students who have used SPARQLuedo and OWLuedo demonstrates the effectiveness of this playful approach for acquiring practical skills in SPARQL and OWL.
