Abstract
Advancements in deep learning have revolutionized materials discovery by enabling predictive modeling of complex material properties. However, single-modal approaches often fail to capture the intricate interplay of compositional, structural, and morphological characteristics. This study introduces a novel multimodal deep learning framework for enhanced material property prediction, integrating textual (chemical compositions), tabular (structural descriptors), and image-based (2D crystal structure visualizations) modalities. Utilizing the Alexandriadatabase, we construct a comprehensive multimodal dataset of 10,000 materials with symmetry-resolved crystallographic data. Specialized neural architectures, such as FT-Transformer for tabular data, Hugging Face Electra-based model for text, and TIMM-based MetaFormer for images, generate modality-specific embeddings, fused through a hybrid strategy into a unified latent space. The framework predicts seven critical material properties, including electronic (band gap, density of states), thermodynamic (formation energy, energy above hull, total energy), magnetic (magnetic moment per volume), and volumetric (volume per atom) features, many governed by crystallographic symmetry. Experimental results demonstrated that multimodal fusion significantly outperforms unimodal baselines. Notably, the bimodal integration of image and text data showed significant gains, reducing the Mean Absolute Error for band gap by approximately 22.7% and for volume per atom by 22.4% compared to the average unimodal models. This combination also achieved a 28.4% reduction in Root Mean Squared Error for formation energy. The full trimodal model (tabular + images + text) yielded competitive, and in several cases the lowest, error metrics, particularly for band gap, magnetic moment per volume and density of states per atom, confirming the value of integrating all three modalities. This scalable, modular framework advances materials informatics, offering a powerful tool for data-driven materials discovery and design.

Abstract
In the age of rapidly advancing machine learning capabilities, the pursuit of maximum performance encounters the practical limitations imposed by limited resources in several fields. This work presents a cost-effective proposal for feature selection, which is a crucial part of machine learning processes, and intends to partly solve this problem through computational time reduction. The proposed methodology aims to strike a careful balance between feature exploration and strict computational time concerns, by enhancing the quality and relevance of data. This approach focuses on the use of interim representations of feature combinations to significantly speed up a potentially slow and computationally expensive process. This strategy is evaluated in several datasets against other feature selection methods, and the results indicate a significant reduction in the temporal costs associated with this process, achieving a mean percentage decrease of 85%. Furthermore, this reduction is achieved while maintaining competitive model performance, demonstrating that the selected features remain effective for the learning task. These results emphasize the method's feasibility, confirming its ability to transform machine learning applications in environments with limited resources.

Abstract
The present study analyzes the types of interrogative, such as yes/no questions, wh-questions, or alternative questions, and their semantic and pragmatic functions in a multilingual parallel corpus of spoken monologues extracted from the TED Talks transcripts in five languages: English, European Portuguese, Lithuanian, Bulgarian, and Hebrew. The corpus was developed with English as the pivot language, and the examples are aligned in all five languages based on the occurrence of an interrogative. To conduct this study, we designed an annotation scheme that harmonizes two parts of ISO 24617 - Part 8: Semantic relations and Part 2: Dialogue acts. This framework enabled us to determine the discourse relations that questions establish with the segments to which they are connected and that precede them and their communicative function. In our analysis, we observed that, despite the monologic nature of the corpus, interrogatives are very frequent and diversified across the five languages. Our findings also reveal that the questions are mostly used with a pragmatic function and that the range of discourse relations is less varied. Additionally, the analysis disclosed some pertinent differences between the five languages concerning the translation choices.

Abstract
This position paper presents A4FN, an Agentic Artificial Intelligence (AI) architecture for intent-driven automation in Flying Networks (FNs) using Unmanned Aerial Vehicles (UAVs) as access nodes. A4FN leverages Generative AI and Large Language Models (LLMs) to enable real-time, context-aware network control via a distributed agentic system. It comprises two components: the Perception Agent (PA), which semantically interprets multimodal input - including imagery, audio, and telemetry data - from UAV-mounted sensors to derive Service Level Specifications (SLSs); and the Decision-and-Action Agent (DAA), which reconfigures the network based on inferred intents. A4FN embodies key properties of Agentic AI, including autonomy, goal-driven reasoning, and continuous perception-action cycles. Designed for mission-critical, infrastructure-limited scenarios such as disaster response, it supports adaptive reconfiguration, dynamic resource management, and interoperability with emerging wireless technologies. The paper details the A4FN architecture, its core innovations, and open research challenges in multi-agent coordination and Agentic AI integration in next-generation FNs.

Abstract
Around 80% of websites change significantly or disappear altogether after the first year, resulting in the loss of invaluable information. In this volatile scenario, preserving online content is increasingly essential. This is especially critical for local news outlets, which produce a wealth of information within the unique context of their communities but often lack sufficient archiving resources. In this paper, we take a significant step forward by leveraging the information preserved by the Portuguese Web Archive, Arquivo.pt, to recreate the website of a local news outlet. This online demo grants users direct access to previously lost news articles, images, and front covers, thus contributing to preserving local digital heritage. An IR system was also implemented to ensure easy access, along with a recommendation system based on BERT embeddings to suggest related news articles and enhance user engagement. As a final contribution, we also provide a Python package, enabling others to replicate the process of collecting, processing, retrieving, and recreating websites for local news outlets in Portugal.

Abstract
The electrification of transportation, driven by the widespread adoption of electric vehicles and increased integration of renewable energy, is critical to decarbonizing mobility and society. Demand response strategies, such as dynamic pricing, enable indirect control of charging processes, but their success relies on accurately estimating consumer responses to tariff changes. Observational data can provide insights into consumer behavior, but the presence of confounding variables motivates the use of causal inference techniques for a reliable elasticity estimation. This study proposes a data-driven framework for optimizing day-ahead charging tariffs, leveraging causal discovery and inference algorithms validated on a synthetically generated dataset. A sensitivity analysis explores the impact of data availability on elasticity estimation and the performance of the resulting demand response strategy. The findings highlight the potential of causal machine learning to characterize consumers and, ultimately, the need for regular characterization to improve the efficiency of demand-side management.