Abstract
We present PolyNarrative, a new multilingual dataset of news articles, annotated for narratives. Narratives are overt or implicit claims, recurring across articles and languages, promoting a specific interpretation or viewpoint on an ongoing topic, often propagating mis/disinformation. We developed two-level taxonomies with coarse- and fine-grained narrative labels for two domains: (i) climate change and (ii) the military conflict between Ukraine and Russia. We collected news articles in four languages (Bulgarian, English, Portuguese, and Russian) related to the two domains and manually annotated them at the paragraph level. We make the dataset publicly available, along with experimental results of several strong baselines that assign narrative labels to news articles at the paragraph or the document level. We believe that this dataset will foster research in narrative detection and enable new research directions towards more multi-domain and highly granular narrative related tasks.

Abstract
Building on theories of human sound perception and spatial cognition, this paper introduces a sonification method that facilitates navigation by auditory cues. These cues help users recognize objects and key urban architectural elements, encoding their semantic and spatial properties using non-speech audio signals. The study reviews advances in object detection and sonification methodologies, proposing a novel approach that maps semantic properties (i.e., material, width, interaction level) to timbre, pitch, and gain modulation and spatial properties (i.e., distance, position, elevation) to gain, panning, and melodic sequences. We adopt a three-phase methodology to validate our method. First, we selected sounds to represent the object’s materials based on the acoustic properties of crowdsourced annotated samples. Second, we conducted an online perceptual experiment to evaluate intuitive mappings between sounds and object semantic attributes. Finally, in-person navigation experiments were conducted in virtual reality to assess semantic and spatial recognition. The results demonstrate a notable perceptual differentiation between materials, with a global accuracy of .69 ± .13 and a mean navigation accuracy of .73 ± .16, highlighting the method’s effectiveness. Furthermore, the results suggest a need for improved associations between sounds and objects and reveal demographic factors that are influential in the perception of sounds.

Abstract
Cities worldwide have agreed on ambitious goals regarding carbon neutrality. To do so, policymakers seek ways to foster smarter and cleaner transportation solutions. However, citizens lack awareness of their carbon footprint and of greener mobility alternatives such as public transports. With this, three main challenges emerge: (i) increase users’ awareness regarding their carbon footprint, (ii) provide personalized recommendations and incentives for using sustainable transportation alternatives and, (iii) guarantee that any personal data collected from the user is kept private. This paper addresses these challenges by proposing a new methodology. Created under the FranchetAI project, the methodology combines federated Artificial Intelligence (AI) and Greenhouse Gas (GHG) estimation models to calculate the carbon footprint of users when choosing different transportation modes (e.g., foot, car, bus). Through a mobile application that keeps the privacy of users’ personal information, the project aims at providing detailed reports to inform citizens about their impact on the environment, and an incentive program to promote the usage of more sustainable mobility alternatives.

Abstract
In many organizations with a distributed operation, not only is data collection distributed, but models are also developed and deployed separately. Understanding the combined knowledge of all the local models may be important and challenging, especially in the case of a large number of models. The automated development of consensus models, which aggregate multiple models into a single one, involves several challenges, including fidelity (ensuring that aggregation does not penalize the predictive performance severely) and completeness (ensuring that the consensus model covers the same space as the local models). In this paper, we address the latter, proposing two measures for geometrical and distributional completeness. The first quantifies the proportion of the decision space that is covered by a model, while the second takes into account the concentration of the data that is covered by the model. The use of these measures is illustrated in a real-world example of academic management, as well as four publicly available datasets. The results indicate that distributional completeness in the deployed models is consistently higher than geometrical completeness. Although consensus models tend to be geometrically incomplete, distributional completeness reveals that they cover the regions of the decision space with a higher concentration of data.

Abstract
Electronic Health Records (EHRs) contain vast amounts of unstructured narrative text, posing challenges for organization, curation, and automated information extraction in clinical and research settings. Developing effective annotation schemes is crucial for training extraction models, yet it remains complex for both human experts and Large Language Models (LLMs). This study compares human- and LLM-generated annotation schemes and guidelines through an experimental framework. In the first phase, both a human expert and an LLM created annotation schemes based on predefined criteria. In the second phase, experienced annotators applied these schemes following the guidelines. In both cases, the results were qualitatively evaluated using Likert scales. The findings indicate that the human-generated scheme is more comprehensive, coherent, and clear compared to those produced by the LLM. These results align with previous research suggesting that while LLMs show promising performance with respect to text annotation, the same does not apply to the development of annotation schemes, and human validation remains essential to ensure accuracy and reliability. © 2025 Copyright for this paper by its authors.

Abstract
Music therapy has emerged as a promising approach to support various mental health conditions, offering non-pharmacological therapies with evidence of improved well-being. Rapid advancements in artificial intelligence (AI) have recently opened new possibilities for ‘personalized’ musical interventions in mental health care. This article explores the application of AI in the context of mental health, focusing on the use of machine learning (ML), deep learning (DL), and generative music (GM) to personalize musical interventions. The methodology included a scoping review in the Scopus and PubMed databases, using keywords denoting emerging AI technologies, music-related contexts, and application domains within mental health and well-being. Identified research lines encompass the analysis and generation of emotional patterns in music using ML, DL, and GM techniques to create musical experiences adapted to user needs. The results highlight that these technologies effectively promote emotional and cognitive well-being, enabling personalized interventions that expand mental health therapies.