Nuno Silva

Abstract
Hierarchical document classification is essential for structuring large-scale textual corpora in domains such as digital libraries and academic repositories. While recent advances in large language models (LLMs) have opened new possibilities for text classification, their applicability to hierarchical settings under real-world constraints remains underexplored. This study investigates both generative and discriminative transformer-based models, evaluating their effectiveness across multiple inference strategies: zero-shot baseline, local fine-tuning, and a global approach using category-specific models. Experiments on two real-world hierarchical datasets provide a comprehensive comparison of classification accuracy, F1-macro scores, and inference times. The results highlight that, although generative LLMs can deliver competitive (yet variable) performance at higher levels of the hierarchy, their high inference costs hinder their use in time-sensitive applications. In contrast, fine-tuned discriminative models—particularly BERT-based architectures—consistently offer a more favorable trade-off between performance and efficiency. © 2025 Elsevier B.V., All rights reserved.

Abstract
uebe.Q is a managing software for solid referential information systems, such as ISO 9000 (for quality) and ISO 1400 (for environment). This is a long-term developed software, encompassing extensive and solid business logic with a long and successful record of deployments. A recent business model change imposed that the evolution and configuration of the software, shifts from the company (and especially the development team) to consultants and other business partners, along with the fact that different systems and respective data/information need to be integrated with minimal intervention of the development team. The so far acceptable rigidity, fragility, immobility and opacity of the software became a problem. Especially, the system was prepared to deal with a specific database respecting a specific schema and code-defined semantics. This paper describes the approach taken to overcome the problems derived form the previous architecture, by adopting (i) ontologies for the specification of business concepts and (ii) an information-integration Decision Support System (DSS) for mapping the domain specific ontologies to the database schemas. © 2020, Springer Nature Switzerland AG.

Abstract
For the last decade, uebe.Q is being adopted by companies in different business areas and countries for managing compliance with solid referential information systems, such as ISO 9000 (for quality) and ISO 1400 (for environment). This is a long-term developed software, encompassing extensive, solid and valuable business logic. When it is deployed for/on a company, it usually demands an extensive and specific adaptation (i.e. software refinement) and configuration process involving DigitalWind's ISO 9000 and ISO 1400 experts as well as software development and operation teams. However, a recent business model change imposed that the evolution and configuration of the software, shifts from DigitalWind (and especially from the development team) to external consultants and to other business partners, along with the fact that different third-party's systems and respective data/information need to be integrated with minimal intervention of the development team. This paper presents and overview of the re-engineering process taken to handle this business model change by adopting (i) ontologies for the specification of business concepts, (ii) closed-world assumption (CWA) rules for the specification of the dynamics of the system and (iii) Domain Specific Language (DSL) for the configuration of the system by domain/business experts. The DSL approach is further described in detail. © 2018 IEEE.

Abstract
Analysis of automatically-generated alignments shows that ambiguous situations are quite common, preventing their application in scenarios demanding high quality and completeness, such ontology mediation (e.g. data transformation and information/knowledge integration). Even the best-performing alignment needs to be manually corrected, completed and verified before application. In this paper, we propose a decision support system (DSS) based in a general-purpose rule engine that assists the expert on improving and completing the automatically-generated alignments into fully-fledged alignments, balancing the precision and recall of the system with the user participation in the process. For that, the rules capture the preconditions (existing facts) and the actions to solve each (ambiguous) alignment scenario, in which the expert decision will be adopted in further automatic decisions. The evaluation of the proposed DSS shows the gain in reducing the need for expert's decisions while increasing the accuracy of the alignments. © 2017 Copyright is held by the owner/author(s).

Abstract
Ontology matchers establish correspondences between ontologies to enable knowledge from different sources and domains to be used in ontology mediation tasks (e.g. data transformation and information/ knowledge integration) in many ways. While these processes demand great quality alignments, even the best-performing alignment needs to be corrected and completed before application. In this paper, we propose a rule-based system that improves and completes the automatically-generated alignments into fullyfledged alignments. For that, the rules capture the pre-conditions (existing facts) and the actions to solve each (ambiguous) scenario, in which automatic decisions supported by a folksonomy-based matcher are adopted. The evaluation of the proposed system shows the increasing accuracy of the alignments.