Portail HAL Nantes Université
Not a member yet
120169 research outputs found
Sort by
Image incomplète : Une étude d'état de l'art sur les biais dans les grands modèles de langage
No full textNational audienceLes grands modèles de langage (LLM) pré-entraînés ont transformé le traitement du langage naturel (TALN) et les tâches quotidiennes, surpassant les méthodes traditionnelles. Leur utilisation a démocratisé l'accès, facilitant l'écriture, le codage et les conseils de santé. Entraînés sur d'immenses corpus textuels issus d'internet, les LLM héritent de biais, perpétuant des stéréotypes qui peuvent fausser les représentations linguistiques et causer des préjudices représentationnels ou allocationnels. Dans le domaine médical, où les LLM soutiennent la communication et la documentation, ces biais présentent des risques significatifs. Cette revue analyse les recherches sur les biais des LLM, identifie les lacunes concernant les déterminants sociaux de la santé (DSS) et discute de la nécessité d'un cadre pour les aborder de manière exhaustive, améliorant l'intégration sécurisée des LLM en santé
Summarization for Generative Relation Extraction in the Microbiome Domain
No full textNational audienceWe explore a generative relation extraction (RE) pipeline tailored to the study of interactions in the intestinal microbiome, a complex and low-resource biomedical domain. Our method leverages summarization with large language models (LLMs) to refine context before extracting relations via instruction-tuned generation. Preliminary results on a dedicated corpus show that summarization improves generative RE performance by reducing noise and guiding the model. However, BERT-based RE approaches still outperform generative models. This ongoing work demonstrates the potential of generative methods to support the study of specialized domains in low-resources setting
Quels sont les bienfaits de l’activité physique dans la gonarthrose ?
No full textInternational audienceKnee osteoarthrosis is a very common multifactorial disease due to the aging population, as well as microtrauma and trauma caused by sports. Physical activities have been a treatment option for knee osteoarthrosis for several years, with the aim of increasing patients’ independence and quality of life while reducing pain and improving joint function. In an attempt to demonstrate the benefits of physical activity in knee osteoarthrosis, this narrative review aimed firstly to verify the significant effectiveness of physical activities, secondly to understand how physical activities can impact the knee joint, and thirdly to identify barriers of physical activity for patients and therapists in order to ultimately present an adaptation of the European recommendations.L’arthrose du genou représente une maladie d’origine multifactorielle très fréquente du fait du vieillissement de la population, mais également en raison des microtraumatismes et des traumatismes occasionnés par la pratique sportive. Les activités physiques font partie depuis quelques années des moyens thérapeutiques de l’arthrose du genou, avec pour objectif d’augmenter l’autonomie et la qualité de vie des patients tout en diminuant les douleurs et améliorant la fonction articulaire. Afin de tenter de démontrer les bienfaits des activités physiques dans la gonarthrose, cette revue narrative a eu pour objet premièrement de vérifier si l’efficacité des activités physiques était significative, deuxièmement de comprendre comment les activités physiques pouvaient avoir une action sur l’articulation du genou, et troisièmement d’identifier les barrières à l’activité physique des patients et thérapeutes, afin de présenter au final une adaptation des recommandations des sociétés savantes européennes
Deciphering the Role of Oxygen in Materials for Heterogeneous Catalysis and Energy Storage: A Dive into the Oxygen K-Edge
No full textInternational audienceUnderstanding the structural and electronic properties of oxygen in materials is critical for advancing various technological applications, particularly in catalysis and energy storage. Oxygen species play a pivotal role in mediating reactions within metal-oxide systems, significantly influencing the reactivity and stability of catalysts. Insight into the electronic structure of oxygen, comprising hybridization and bonding with transition metals, provides a deeper understanding of catalytic processes and the efficiency of electrocatalysts. Additionally, oxygen’s involvement in electrochemical processes within batteries, particularly through anionic redox reactions, underscores its importance in developing efficient energy storage solutions. This review focuses on the use of two empirical parameters in O K-edge simulations, offering a guide on simulating observed O spectra for solid oxides. By bridging experimental fingerprints with theoretical approaches, an in-depth understanding of the spectra is provided. First, the theory defining the empirical parameters of the FDMNES code, “dilatorb” and “screening,” is presented. Then, the importance of studying the O K-edge is highlighted through a discussion of selected case studies. This highlights the essential role of oxygen K-edge experiments and simulations in advancing the understanding of catalytic and energy storage processes. By providing detailed insights into the electronic structure, bonding, and real-time transformations in (electro)catalysts, these techniques make significant contributions to material design and optimization
New Insights Into Renal Involvement During Immune-Mediated Thrombotic Thrombocytopenic Purpura
No full textInternational audienc
Attention-Aware Graph Neural Network Modeling for AIS Reception Area Prediction
No full textInternational audienceAccurately predicting the reception area of the Automatic Identification System (AIS) is critical for ship tracking and anomaly detection, as errors in signal interpretation may lead to incorrect vessel localization and behavior analysis. However, traditional propagation models, whether they are deterministic, empirical, or semi-empirical, face limitations when applied to dynamic environments due to their reliance on detailed atmospheric and terrain inputs. Therefore, to address these challenges, we propose a data-driven approach based on graph neural networks (GNNs) to model AIS reception as a function of environmental and geographic variables. Specifically, inspired by attention mechanisms that power transformers in large language models, our framework employs the SAmple and aggreGatE (GraphSAGE) framework convolutions to aggregate neighborhood features, then combines layer outputs through Jumping Knowledge (JK) with Bidirectional Long Short-Term Memory (BiLSTM)-derived attention coefficients and integrates an attentional pooling module at the graph-level readout. Moreover, trained on real-world AIS data enriched with terrain and meteorological features, the model captures both local and long-range reception patterns. As a result, it outperforms classical baselines—including ITU-R P.2001 and XGBoost in F1-score and accuracy. Ultimately, this work illustrates the value of deep learning and AIS sensor networks for the detection of positioning anomalies in ship tracking and highlights the potential of data-driven approaches in modeling sensor reception
Dark‐Toned Halite‐Enriched Veins Above the Marker Band Record a Drying Environment in Gale Crater
No full textInternational audienceThe Martian surface preserves evidence of a global climate transition from wetter to drier conditions, but the nature of the fluids involved in this evolution remains poorly constrained. In Gale crater, the clay-sulfate transition and presence of evaporite mineral assemblages can provide insights into the properties of these fluids and the timing of environmental change. While traversing through the Chenapau member of the sulfate-bearing unit in Gale crater, the Curiosity rover encountered a set of dark-toned veins enriched in Na and Cl, suggestive of halite. However, previous halite detections in Gale crater have been limited to occurrences along the edges of Ca-sulfate veins or nodules, suggesting a unique origin for this set of veins. Here, we hypothesize that these veins formed through the infiltration of saline fluids along pre-existing hydraulically induced fractures. These fluids permeated into the host rock beyond the primary fractures, precipitating halite and cementing the fractures. Using Mastcam and ChemCam spectra, we found that the veins displayed a downturn in the near-infrared wavelengths, consistent with the presence of ferrous iron. Furthermore, textural analysis of the veins reveals host rock material preserved within the veins. ChemCam laser-induced breakdown spectroscopy observations also support the presence of a minor Fe component in the veins and halite concentrated along the center of the fractures. Our results demonstrate that these veins represent a distinct class of diagenetic features in Curiosity's mission that record an important transition in near-surface fluid chemistry consistent with a transition to a drier environment.
Instrumentation of JUNO 3-inch PMTs
No full textInternational audienceOver 25,600 3-inch photomultiplier tubes (PMTs) have been instrumented for the central detector of the Jiangmen Underground Neutrino Observatory. Each PMT is equipped with a high-voltage divider and a frontend cable with waterproof sealing. Groups of sixteen PMTs are connected to the underwater frontend readout electronics via specialized multi-channel waterproof connectors. This paper outlines the design and mass production processes for the high-voltage divider, the cable and connector, as well as the waterproof potting of the PMT bases. The results of the acceptance tests of all the integrated PMTs are also presented
In situ multiscale characterization of cellulose nanocrystals orthotropic organization achieved by combining ultrasound and frontal ultrafiltration
No full textInternational audienceCombining frontal ultrafiltration (FU) and ultrasound (US) processes, an orthotropic organization of cellulose nanocrystals (CNCs) suspensions was achieved and characterized at multiscale, from the nanometer length scale by small angle X-ray scattering (SAXS) and dichroism, to the micrometer length scale by small angle light scattering (SALS). A dedicated channel cell has been developed to simultaneously generate a vertical acoustic force via a vibrating blade at the top and a transmembrane pressure force through the membrane at the bottom. Three specific orientations in three successive layers were revealed in a one-step FU-US process. In a first layer at the bottom, the CNCs were concentrated under the action of transmembrane pressure with their director aligned parallel to the horizontal membrane surface. In a second upper layer a random orientation of the CNCs was detected. Finally, in the upper part of the channel, a third layer of CNCs with their director vertically oriented was revealed. Near the membrane surface, transmembrane pressure forces dominate, leading to highly concentrated CNCs deposition with enough consistency to avoid any change in orientation induced by the US at the top of the channel. At increasing distances from the membrane, acoustic radiation forces become predominant, reorienting the CNCs vertically
