Motivated by the astonishing capabilities of large language models (LLMs) in text-generation, reasoning, and simulation of complex human behaviors, in this paper, we propose a novel multi-component LLM-based framework, namely LLM4ACOE, that fully automates the collaborative ontology engineering (COE) process using role-playing simulation of LLM agents and retrieval augmented generation (RAG) technology. The proposed solution enhances the LLM-powered role-playing simulation with RAG ‘feeding’ the LLM with three different types of external knowledge. This knowledge corresponds to the knowledge required by each of the COE roles (agents), using a component-based framework, as follows: (a) domain-specific data-centric documents, (b) OWL documentation, and (c) ReAct guidelines. The aforementioned components are evaluated in combination, with the aim of investigating their impact on the quality of generated ontologies. The aim of this work is twofold, (a) to identify the capacity of LLM-based agents to generate acceptable (by human-experts) ontologies through agentic collaborative ontology engineering (ACOE) role-playing simulation, at specific levels of acceptance (accuracy, validity, and expressiveness of ontologies) without human intervention and (b) to investigate whether and/or to what extent the selected RAG components affect the quality of the generated ontologies. The evaluation of this novel approach is performed using ChatGPT-o in the domain of search and rescue (SAR) missions. To assess the generated ontologies, quantitative and qualitative measures are employed, focusing on coverage, expressiveness, structure, and human involvement.

The article is concerned with realizability in abstract argumentation. It provides characterization theorems for the most basic types of labelling-based semantics, namely conflict-free and naive labellings. It turns out that existing characterizations for extension-based semantics are of little help in characterizing labelling-based semantics. To this end, we introduce several new criteria like L-tightness, reject-witnessing, reject-compositionality as well as the new construct of a labelling-downward-closure, which help determine whether a given set of labellings is realizable regarding conflict-free or naive semantics. Moreover, we present standard constructions and analyse their uniqueness status. Further classical concepts like ordinary and strong equivalence are studied too. Last but not least, we delve into the characterization of stable labellings. It turns out that this endeavour is a highly non-trivial task with many parallels to so-called compact realizability, an open problem for stable semantics in abstract argumentation.

We consider nonmonotonic inferences from belief bases that contain conditionals enforcing some of the possible worlds to be infeasible and thus completely implausible. In contrast to belief bases satisfying the strong notion of consistency requiring every world to be at least somewhat plausible, we call such belief bases weakly consistent. First, we review the treatment of weakly consistent belief bases by the seminal approaches of p-entailment, which coincides with system P, and of system Z, which coincides with rational closure. Then we focus on c-inference, an inductive inference operator that has been shown to exhibit many desirable properties put forward for nonmonotonic reasoning. It is based on c-representations, which are a special kind of ranking model ordering worlds according to their plausibility. While c-representation is defined for strongly consistent belief bases only, in this article, we extend the notions of c-representation and of c-inference to cover also weakly consistent belief bases. We adapt a constraint satisfaction problem (CSP) characterizing c-representations to capture extended c-representations, and we show how this extended CSP can be used to characterize extended c-inference, providing a basis for its implementation. We show various properties of extended c-inference and in particular, we prove that also the extended notion of c-inference fully satisfies syntax splitting. Furthermore, we extend and evaluate credulous and weakly skeptical c-inference to weakly consistent belief bases and provide characterizations for them as CSPs.

Plan repair is the problem of solving a given planning problem by using a solution plan of a similar problem. This paper presents the first approach where the repair has to be done optimally, that is, we aim at finding a minimum distance plan from an input plan; we do so by introducing a number of compilation schemes that convert a classical planning problem into another where optimal plans correspond to plans with the minimum distance from an input plan. We also address the problem of finding a minimum distance plan from a set of input plans, instead of just one plan. Our experiments using a number of planners show that such a simple approach can solve many problems optimally and more effectively than replanning from scratch for a large number of cases. Also, the approach proves competitive with ${\mathsf{LPG}\textrm{-}\mathsf{adapt}}$, a state-of-the-art approach for the plan repair problem.

One common approach to solve multi-objective reinforcement learning (MORL) problems is to extend conventional Q-learning by using vector Q-values in combination with a utility function. However issues can arise with this approach in the context of stochastic environments, particularly when optimising for the scalarised expected reward (SER) criterion. This paper extends prior research, providing a detailed examination of the factors influencing the frequency with which value-based MORL Q-learning algorithms learn the SER-optimal policy for an environment with stochastic state transitions. We empirically examine several variations of the core multi-objective Q-learning algorithm as well as reward engineering approaches and demonstrate the limitations of these methods. In particular, we highlight the critical impact of the noisy Q-value estimates issue on the stability and convergence of these algorithms.

While modern definitions of business processes exist and are shared in the business process management (BPM) community, a commonly agreed meta-model is still missing. Nonetheless, several different business process meta-models have been proposed and discussed in the literature, which look at business process models from different perspectives, focusing on different aspects and often using different labels for denoting the same element or element relation.

In this paper, we extend and consolidate an effort of building a business process meta-model starting from elements and relations discovered inspecting relevant literature through a systematic literature review. The obtained literature-based business process meta-model, which is on purpose built to disclose critical issues, is then inspected, compared to a previous, more restricted, version, and discussed. The analysis confirms a lack of attention to some crucial business process elements, as well as the presence of some unclear relations and subsumption cycles. Moreover it brings about new issues and inconsistencies in the meta-models proposed in literature, which we address - at least in part - using an ontological analysis.

Exception handling has been successfully proposed in the past years as a simple yet powerful software engineering tool to promote modularity and decoupling, while also preserving robustness. Multi-agent systems (MAS) and organizations (MAOs), in turn, offer powerful abstractions to build distributed systems; current models and methodologies, however, fall short in addressing exception handling in a systematic way, not considering exceptions as part of their design. Thus, the problem is usually approached by ad hoc solutions that hamper code modularization and decoupling. In this work, we outline a vision of how exception handling in MAS can be granted by design. We present an extension of the organizational model and infrastructure adopted in JaCaMo, that explicitly encompasses the notion of exception as a first-class element in the design of an organization. Relying on such a model, we propose an exception handling mechanism that is seamlessly integrated with organizational concepts, such as responsibilities, goals, and norms. In an organization, besides responsibilities for organizational goals, we propose to specify also responsibilities for managing exceptions, that is, for providing feedback about the context in which exceptions occur, and for handling it.

Advertising click-through rate (CTR) prediction is a fundamental task in recommender systems, aimed at estimating the likelihood of users interacting with advertisements based on their historical behavior. This prediction process has evolved through two main stages: from traditional shallow interaction models to more advanced deep learning approaches. Shallow models typically operate at the level of individual features, failing to fully leverage the rich, multilevel information available across different feature sets, leading to less accurate predictions. In contrast, deep learning models exhibit superior feature representation and learning capabilities, enabling a more realistic simulation of user interactions and improving the accuracy of CTR prediction. This paper provides a comprehensive overview of CTR prediction algorithms in the context of recommender systems. The algorithms are categorized into two groups: shallow interactive models and deep learning-based prediction models, including deep neural networks, convolutional neural networks, recurrent neural networks, and graph neural networks. Additionally, this paper also discusses the advantages and disadvantages of the aforementioned algorithms, as well as the benchmark datasets and model evaluation methods used for CTR prediction. Finally, it identifies potential future research directions in this rapidly advancing field.

As the global population ages, effective home healthcare solutions become essential. Over a decade ago, ambient-assisted living (AAL) emerged as a promising solution, especially when combined with the potential of the Internet of Things (IoT) to revolutionize healthcare delivery. However, integrating diverse smart home devices with healthcare systems poses challenges regarding interoperability and real-time, context-aware responses. Addressing these challenges, this study introduces an ontology for AAL that seamlessly merges IoT and Smart Home ontologies with the established healthcare ontology, SNOMED CT. This ontology-centric approach facilitates semantic interoperability and knowledge sharing, paving the way for more personalized healthcare delivery. The core of this work lies in developing an AAL monitoring system grounded in this ontology. By incorporating Semantic Web Rule Language (SWRL) rules, the system can provide context-sensitive automated alerts and responses, taking into account patient-specific attributes, household features, and instantaneous sensor data. Empirical testing in the Halmstad Intelligent Home (HINT) highlights the system’s viability for practical deployment. Preliminary results indicate that the proposed integrative ontology-driven strategy holds significant potential to enhance healthcare services in AAL environments, marking an essential step towards achieving personalized, patient-centric care.

Traditional frequent itemset mining (FIM) is constrained by several limitations, mainly due to its failure to account for item quantity and significance, including factors such as price and profit. To address these limitations, high utility itemset mining (HUIM) is presented. Traditional HUIM algorithms are designed to operate solely on static transactional datasets. Nevertheless, in practical applications, datasets tend to be dynamic, with examples like market basket analysis and business decision-making involving regular updates to the data. Dynamic datasets are updated incrementally with the frequent addition of new data. Incremental HUIM (iHUIM) approaches mine the high utility itemsets (HUIs) from incremental datasets without scanning the whole dataset. In contrast, traditional HUIM approaches require a full dataset scan each time the dataset is updated. Consequently, iHUIM approaches effectively reduce the computational cost of identifying HUIs whenever a new record is added. This survey provides a novel taxonomy that includes two-based, pattern-growth-based, projection-based, utility-list-based, and pre-large-based algorithms. The paper delivers an in-depth analysis, covering the features and characteristics of the existing state-of-the-art algorithms. Additionally, it supplies a detailed comparative overview, advantages, disadvantages, and future research directions of these algorithms. The survey provides both a categorized analysis and a comprehensive, consolidated summary and analysis of all current state-of-the-art iHUIM algorithms. It offers a more in-depth comparative analysis than the currently available state-of-the-art surveys. Additionally, the survey highlights several research opportunities and future directions for iHUIM.