Abstract
Monodic folk music has traditionally been preserved in physical documents. It constitutes a vast archive that needs to be digitized to facilitate comprehensive analysis using AI techniques. A critical component of music score digitization is the transcription of lyrics, an extensively researched process in Optical Character Recognition (OCR) and document layout analysis. These fields typically require the development of specific models that operate in several stages: first, to detect the bounding boxes of specific texts, then to identify the language, and finally, to recognize the characters. Recent advances in vision language models (VLMs) have introduced multimodal capabilities, such as processing images and text, which are competitive with traditional OCR methods. This paper proposes an end-to-end system for extracting lyrics from images of handwritten musical scores. We aim to evaluate the performance of two state-of-the-art VLMs to determine whether they can eliminate the need to develop specialized text recognition and OCR models for this task. The results of the study, obtained from a dataset in a real-world application environment, are presented along with promising new research directions in the field. This progress contributes to preserving cultural heritage and opens up new possibilities for global analysis and research in folk music.

Abstract
Supervised Machine Learning (ML) models require large amounts of labeled data for training. However, this becomes challenging when dealing with resource- and network-constrained Internet of Things (IoT) devices that collect data. Furthermore, in scenarios where the acquired data is fast-changing and highly temporal, continuous and online learning becomes necessary. In this paper, we address the problem of efficiently training ML models using data from IoT nodes. We specifically focus on two aspects: i) selecting the nodes that provide data for the re/training, and ii) determining the optimal amounts of data to be acquired from these nodes, considering network and time constraints, while minimizing learning errors. To tackle this optimization problem, we propose ONDS: an Optimum Node and Data Selection algorithm with linear complexity in the worst-case. ONDS offers a model-agnostic solution applicable to different data modalities and ML architectures. To evaluate the performance of ONDS, we conduct experiments using various models and real-world datasets. The results demonstrate the effectiveness of ONDS, as it outperforms existing alternatives in both classification and regression tasks.

Abstract
Ultra-wideband (UWB) systems are critical for indoor positioning in robotics, industrial tracking, and asset management due to their accuracy in multipath-prone environments. Like GPS satellites requiring precise orbital data, UWB systems depend on well-calibrated anchors-fixed reference points whose positional accuracy directly impacts location estimates. We systematically evaluate and compare computational calibration methods, such as Genetic Algorithms, Maximum Likelihood, and the Extended Kalman Filter, using synthetic data, assessing both efficiency and error reduction in calibration and location. Nonlinear Least Squares (NLS) outperformed other approaches from this review as well as state-of-the-art methods, reducing anchor calibration errors to 10.7cm (86.03% improvement from 1-meter initial uncertainty) and tag localization errors to 5.6cm (88.35% reduction). NLS maintained computational efficiency (mean execution time of 0.011s, proving ideal for real-world deployments where efficiency and accuracy are critical.

Abstract
Rising temperatures, extreme precipitation events such as excessive or insufficient rainfall, increasing levels of carbon dioxide, and associated climatic factors will persistently impact crop growth and agricultural production. The warming temperatures have reduced the agricultural crop yields. Rice (Oryza sativa L.) is the major food crop, which is particularly susceptible to the effects of climate change. It is very important to accurately evaluate the impacts of climate change on rice growth and rice yield. In this study, the rice growth during 1981-2018 (baseline period) and 2041-2100 (future period) were separately simulated and compared within the CERES-Rice model (v4.6) using high-quality weather data, soil, and field experimental data at six agro-meteorological stations in Hainan Province. For the climate data of the future period, the SSP1-2.6, SSP3-7.0, and SSP5-8.5 scenarios were applied, with carbon dioxide (CO2) fertilization effects considered. The adaptation strategies such as adjusting planting dates and switching rice cultivars were also assessed. The simulation results indicated that the early rice yields in the 2050s, 2070s, and 2090s were projected to decrease by 6.2%, 11.8%, and 20.0% when the CO2 fertilization effect was not considered, compared with the results of the baseline period, respectively, while late rice yields would decline by 9.9%, 23.4%, and 36.3% correspondingly. When accounting for the CO2 fertilization effect, the yields of early rice and late rice in the 2090s increased 16.9% and 6.2%, respectively. Regarding adaptation measures, adjusting planting dates and switching rice cultivars could increase early rice yields by 22.7% and 43.3%, respectively, while increasing late rice yields by 20.2% and 34.2% correspondingly. This study holds substantial scientific importance for elucidating the mechanistic pathways through which climate change influences rice productivity in tropical agro-ecosystems, and provides a critical foundation for formulating evidence-based adaptation strategies to mitigate climate-related risks in a timely manner. Cultivar substitution and temporal shifts in planting dates constituted two adaptation strategies for attenuating the adverse impacts of anthropogenic climate change on rice.

Abstract
Ensuring food quality and safety is a growing challenge in the food industry, where early detection of contamination or spoilage is crucial. Using gas sensors combined with Artificial Intelligence (AI) offers an innovative and effective approach to food identification, improving quality control and minimizing health risks. This study aims to evaluate food identification strategies using supervised learning techniques applied to data collected by the BME Development Kit, equipped with the BME688 sensor. The dataset includes measurements of temperature, pressure, humidity, and, particularly, gas composition, ensuring a comprehensive analysis of food characteristics. The methodology explores two strategies: a neural network model trained using Bosch BME AI-Studio software, and a more flexible, customizable approach that applies multiple predictive algorithms, including DT, LR, kNN, NB, and SVM. The experiments were conducted to analyze the effectiveness of both approaches in classifying different food samples based on gas emissions and environmental conditions. The results demonstrate that combining electronic noses (E-Noses) with machine learning (ML) provides high accuracy in food identification. While the neural network model from Bosch follows a structured and optimized learning approach, the second methodology enables a more adaptable exploration of various algorithms, offering greater interpretability and customization. Both approaches yielded high predictive performance, with strong classification accuracy across multiple food samples. However, performance variations depend on the characteristics of the dataset and the algorithm selection. A critical analysis suggests that optimizing sensor calibration, feature selection, and consideration of environmental parameters can further enhance accuracy. This study confirms the relevance of AI-driven gas analysis as a promising tool for food quality assessment.

Abstract
The Eighth International Workshop on Narrative Extraction from Texts (Text2Story'25) was held on April 10 th , 2025, in conjunction with the 47 th European Conference on Information Retrieval (ECIR 2025) in Lucca, Italy. During this half-day event, more than 30 attendees engaged in discussions and presentations focused on recent advancements in narrative representation, extraction, and generation. The workshop featured a keynote address and a mix of oral presentations and poster sessions covering nineteen papers. The workshop proceedings are available online 1 . Date: 10 April 2025. Website: https://text2story25.inesctec.pt/.