HAL Université de Tours
Not a member yet
55851 research outputs found
Sort by
A new experimental methodology for swelling and shrinkage assessment of biobased materials
No full textInternational audienceThe high sensitivity of bio-composites to hygrothermal conditions affects their performance in real-life conditions resulting in dimensional variations. The measurement of the displacement resulting from swelling and shrinkage is a subject of great complexity due to their heterogeneity. The objective of this work is to develop an efficient methodology to measure and quantify these dimensional variations. The paper presents a new experimental investigation based on laser displacement sensors. The main feature of these sensors is that they provide nondestructive and non-contact measurements of the sample. The measurements are made after a cure of the samples of 3 months and an exposure to different humidity conditions. The exposure of the samples to variable humidity consists of a cycle of humidification and drying by several levels. This leads to the identification of specific hygroexpansion and shrinkage coefficients β for each formulation. The results show a highlighting of the phenomena of swelling and shrinkage depending on the type of aggregates with non-reversible behavior with significant hysteresis and inverse behaviour according to different directions. This behaviour is due to water sorption and plant type aggregates. Two treatments are then applied to control the swelling and shrinkage behavior of the aggregates in the biocomposite by using NaCl and NaHCO 3 solutions. The treatment of the aggregates decreases the rate of swelling and shrinkage of biocomposites, highlighting a key role of NaHCO 3 treatment.</div
Modélisation et compréhension des dynamiques des relations dans les graphes de connaissances avec des applications au classement et à l'analyse de la stabilité
No full textKnowledge Graphs (KGs) in the Semantic Web have become central to numerous applications in artificial intelligence, semantic search, and data integration. Their collaborative, crowdsourced nature, represented by platforms like Wikidata, DBpedia, and YAGO, offers immense scale and coverage but also raises important questions about their structural dynamics and analytic utility. While much research has focused on ontological (or schema) modeling and reasoning, less attention has been devoted to understanding how facts evolve and self-organize in the factual layer of these graphs. This thesis addresses this gap by modeling the dynamic behavior of relationships in KGs, discovering domain-specific ranking indicators, and assessing the robustness of resulting rankings under perturbations. Our investigation is guided by four research questions. First, we explore whether the distributed editing processes in large crowdsourced KGs result in stable global structures (RQ1). Second, we seek to understand how to model the knowledge accumulation process (RQ2). Building on these insights, we address the problem of automatically discovering meaningful and interpretable ranking indicators tailored to specific domains and use cases (RQ3). Finally, we assess the resilience of such rankings in the face of structural errors and vandalism (RQ4). To tackle these questions, we first introduce KRELM, a generative Knowledge Relationship Model that treats each KG relationship as a bipartite network governed by an asymmetric attachment process: subjects receive new facts uniformly, while popular objects attract new facts preferentially. This model explains the persistent emergence of exponential and power-law degree distributions in real-world KGs. We provide theoretical theorems for these convergence behaviors and empirically validate KRELM across four crowdsourced KGs and eight historical Wikidata snapshots. Our results demonstrate that the structural regularities observed in KGs are not coincidental but are rooted in reproducible generative mechanisms. Building on this foundation, we propose RIPM (Ranking-Indicator Pattern Miner), a scalable algorithm for the automatic extraction of domain specific ranking indicators. RIPM identifies, filters, and favors relationship class pairs based on their statistical inequality (measured by the Gini coefficient) and coverage (the proportion of covered entities). We derive an efficient approximation for the Gini coefficient, enabling RIPM to operate within the constraints of public SPARQL endpoints. An experimental evaluation, including a user study with 19 participants, confirms the utility, diversity, and interpretability of the indicators discovered by RIPM. Finally, we examine the robustness of rankings under structural perturbations by proposing a probabilistic model of ranking stability. We formalize global and local perturbation scenarios and quantify the likelihood that an entity’s rank changes under perturbations. Building upon KRELM’s growth dynamics, we establish theoretical thresholds for tolerable perturbations and empirically validate these results across multiple KGs. Overall, this thesis presents a framework that links the dynamic structure of KGs with knowledge graph analysis, in particular, ranking indicators mining and robustness analysis. By combining insights from complex networks and statistical modeling, it advances the theoretical understanding and practical exploitation of large-scale, crowdsourced knowledge graphs.Les graphes de connaissances du Web sémantique sont devenus centraux pour de nombreuses applications en intelligence artificielle, en recherche sémantique et en intégration de données. Leur nature collaborative et participative, illustrée par Wikidata, DBpedia et YAGO, leur confère une échelle et une couverture remarquables, tout en soulevant des questions essentielles sur leurs dynamiques structurelles et leur utilité pour des tâches d’analyse. Alors que de nombreux travaux se sont concentrés sur la modélisation ontologique et le raisonnement, peu d’attention a été portée à la manière dont les faits évoluent et s’auto-organisent dans la couche factuelle de ces graphes. Cette thèse comble cette lacune en modélisant le comportement dynamique des relations dans les graphes, en découvrant des indicateurs de classement spécifiques et en évaluant la robustesse des classements obtenus face à des perturbations. Notre investigation est guidée par quatre questions de recherche. Premièrement, nous examinons si les processus d’édition distribuée dans les grands graphes collaboratifs conduisent à des structures globales stables (RQ1). Deuxièmement, nous cherchons à modéliser le processus d’accumulation des connaissances (RQ2). Sur cette base, nous abordons la découverte automatique d’indicateurs de classement pertinents et interprétables, adaptés à des domaines et cas d’usage spécifiques (RQ3). Enfin, nous évaluons la résilience de ces classements face aux erreurs structurelles et au vandalisme (RQ4). Pour répondre à ces questions, nous introduisons KRELM (Knowledge Relationship Model), un modèle génératif qui considère chaque relation d’un graphe comme un graphe biparti régi par un processus d’attachement asymétrique : les sujets reçoivent de nouveaux faits uniformément, tandis que les objets populaires attirent préférentiellement. Il explique l’émergence récurrente de distributions stables, de degrés exponentielles pour les sujets, et en loi de puissance pour les objets, observées dans les graphes réels. Nous fournissons des théorèmes théoriques sur ces comportements de convergence et validons empiriquement KRELM sur quatre graphes collaboratifs et huit instantanés historiques de Wikidata. Nos résultats montrent que les régularités structurelles observées ne sont pas fortuites, mais enracinées dans des mécanismes génératifs reproductibles. Sur cette base, nous proposons RIPM (Ranking Indicator Pattern Miner), un algorithme pour l’extraction automatique d’indicateurs de classement. RIPM identifie, filtre et privilégie les paires relation–classe selon leur inégalité statistique (mesurée par le coefficient de Gini) et leur couverture (proportion d’entités concernées). Nous dérivons une approximation efficace du coefficient de Gini, permettant à RIPM d’opérer dans les contraintes des points d’accès SPARQL publics. Une évaluation expérimentale, incluant une étude utilisateur auprès de 19 participants, confirme l’utilité, la diversité et l’interprétabilité des indicateurs découverts. Enfin, nous examinons la robustesse des classements face aux perturbations structurelles via un modèle probabiliste de stabilité des classements. Nous formalisons des scénarios de perturbations globales et locales et quantifions la probabilité qu’un classement d’entité change sous ces perturbations. En nous appuyant sur les dynamiques de croissance de KRELM, nous établissons des seuils théoriques de perturbations tolérables et validons empiriquement ces résultats sur plusieurs graphes de connaissances. Dans l’ensemble, cette thèse présente un cadre reliant la structure dynamique des graphes de connaissances à leur analyse, notamment l’extraction d’indicateurs de classement et l’étude de leur robustesse. En combinant les apports des réseaux complexes et de la modélisation statistique, elle fait progresser la compréhension théorique et l’exploitation pratique des graphes de connaissances collaboratifs à grande échelle
Operating Units in Written Language Performance: Linguistic and Behavioral Perspectives: Introduction to the Special Issue
No full textInternational audienceThis special issue of Journal of Writing Research addresses the fundamental question of performance units in writing: how can we characterize these units, and which theoretical paradigms allow us to describe them? Despite their core role in the writing activity, there is no consensus on the nature of written performance units. In order to progress on this issue, the different articles presented in this special issue shed light on performance units, their description, definition and role in text construction. Different methodological and theoretical approaches, based on behavioral data, with pauses as a central indicator, illustrate how linguistic structures produced in these units constrain written production
«Entre gestion sécuritaire et mobilisations citoyennes : les émeutes londoniennes de 2011 comme révélatrices d'une gouvernance déficiente»
No full textInternational audienc
Hybridization Effects on the Mechanical and Aging Behavior of Elium Acrylic‐Based Flax‐Glass Fiber Composites
No full textInternational audienceThis study examines the hybridization effect of flax–glass fiber‐reinforced Elium acrylic polymer composites, with a particular emphasis on their mechanical properties, viscoelastic behavior, and diffusion kinetics. For this purpose, hybrid composites were fabricated by combining flax and glass fiber laminates with various stacking sequences using vacuum molding and subjected to environmental aging at room temperature. The primary objective was to evaluate the performance of these hybrid composites in comparison to flax fiber laminate composites. The results demonstrate that hybridizing flax fibers with glass fibers significantly mitigates the degradation of mechanical properties induced by aging. Notably, hybrid composites exhibit a significant improvement in both Young's modulus and tensile strength compared to pure flax fiber laminate composites. Dynamic mechanical analysis (DMA) further reveals that hybrid composites achieve marked enhancements in storage modulus and glass transition temperature, indicating improved viscoelastic behavior. Scanning electron microscopy (SEM) analyses performed on post‐tensile and aging tests highlight a reduction in microcrack propagation due to hybridization. These findings underscore the role of glass fibers in enhancing the durability and overall performance of flax fiber composites, making them a promising candidate for applications requiring sustained exposure to aggressive environments
Pioneering Sustainability in Ingredient Design with Eutectic Matrixes: Focus on the Smart selection strategy
No full textInternational audienceThe cosmetics industry is undergoing a profound transformation, with an increasing demand for natural ingredientsand stricter regulatory requirements driving the need for more sustainable extraction solutions. In response tothese challenges, Natural Deep Eutectic Solvents (NaDES) and their broader classification, Mixtures of NaturalCompounds (MiNaCs), have emerged as innovative alternatives to conventional solvents.1 These systems offerunique properties, such as biodegradability, low toxicity, and compatibility with natural metabolite extraction,making them highly attractive for green chemistry applications.However, to be truly green, MiNaCs must fulfill a dual role as ingredient solvents. This means eliminating theenergy- and resource-intensive step of separating solvents from extracted metabolites, which is often a majordrawback of traditional processes. In this context, our laboratory has developed a “Smart Selection Strategy”,2,3 agroundbreaking approach that integrates regulatory, computational, and experimental constraints into the solventdesign process. This strategy is specifically tailored to meet the needs of the cosmetics market, making it avaluable contribution to sustainable chemistry.The Smart Selection Strategy begins with the application of regulatory filters to identify raw materials that arecompatible with target markets. For example, choline chloride, widely used in NaDES formulations, was excludedfrom our study since it is banned in European cosmetics. This step ensures that all potential solvent candidates arecompliant with existing regulations, reducing the risk of downstream obstacles.4,5With a refined Smart Library, we then applied a computational σ-screening approach using the COSMO-RSplatform. This method rapidly assesses the compatibility of raw materials with target compounds based onmolecular surface properties and interaction potentials. Once promising MiNaC formulations were identified,COSMO-RS was further used to simulate solid-liquid equilibrium (SLE) diagrams, to identify the precise molar ratioswhere intermolecular interactions are strongest. These insights guide the experimental phase, significantlyreducing the time, energy, and material resources typically required for solvent design and selection.To demonstrate the effectiveness of our strategy, we focused on the valorization of apolar metabolites fromSpirulina, a microalgae known for its high-value compounds, including polyunsaturated fatty acids (PUFAs) andpigments. From an initial Smart Library of 15 raw materials, our strategy identified four optimal binary MiNaCmixtures. These mixtures were experimentally evaluated to validate the predictions of the σ-screening approach,revealing strong agreement between computational predictions and laboratory results. Notably, our σ-screeningapproach proved to be more effective than traditional activity coefficient-based solvent selection methods. Thisadvantage was particularly evident in the context of regulatory constraints, where the pool of eligible raw materialswas significantly limited. By identifying optimal solvent formulations with greater efficiency, our method reduces theenvironmental and economic costs of developing new MiNaCs.This study highlights the transformative potential of MiNaCs as sustainable solvents for the cosmetics industry. Byeliminating solvent/metabolite separation and streamlining the design process, MiNaCs align with green chemistryprinciples, such as reducing waste, minimizing energy use, and leveraging renewable resources.Our Smart Selection Strategy represents a multidisciplinary and forward-thinking approach to solvent development,demonstrating how scientific rigor and eco-conscious design can address the growing demand for sustainablesolutions in cosmetics. This methodology not only advances the field of green chemistry but also supports theindustry transition to more responsible and efficient practices
Revealing the Nature of Eutectic Solvents: A Synergy Between DSC and Lock-Free NMR DOSY
No full textInternational audienceThe growing focus on sustainability in chemical processes has led to the search for alternative solvents that areeco-friendly, cost-effective, and versatile. Natural Deep Eutectic Solvents (NaDES) have emerged as promisingcandidates, derived from naturally occurring metabolites, offering a way to reduce reliance on volatile organiccompounds while supporting efficient processes. However, the broad use of the NaDES term in scientific literaturecan sometimes obscure important differences in the composition and properties of the mixtures studied.To properly classify a mixture as a NaDES, rigorous experimental characterization, including the establishment of itssolid/liquid phase diagram, is essential to confirm a true deep eutectic point. Without this, labeling a mixture asNaDES risks oversimplification or misrepresentation. Over the past decade, the scientific community has criticizedthe overuse of the NaDES term, leading to the development of alternative labels such as Low TransitionTemperature Mixtures (LTTMs) and Mixtures of Natural Compounds (MiNaCs).(1–4) LTTMs describe systems thatlack crystallinity upon mixture by exhibiting glass transition temperatures, marking them as fundamentally differentfrom true eutectic solvents. MiNaCs, on the other hand, represent a broader category, encompassing a wide rangeof natural metabolite-based compositions without requiring the presence of either a eutectic point or a glasstransition. This nomenclature also tolerates the inclusion of higher water proportions, extending the practical utilityof these systems in green chemistry applications.In this study, we sought to resolve these terminological and scientific ambiguities by implementing an innovativemulti-technique characterization approach. Specifically, we combined viscosity measurements, differential scanningcalorimetry (DSC), and an original lock-free NMR Diffusion-Ordered Spectroscopy (DOSY) strategy. This approachwas applied to two binary mixtures: octanoic acid paired with menthol and octanoic acid combined with 1,3-propanediol. These experimental analyses were complemented by SLE (Solid-Liquid Equilibrium) simulations usingCOSMO-RS, providing a theoretical framework for interpretating of the observed behavior.At a 1:1 molar ratio, the octanoic acid/menthol system exhibited properties consistent with genuine NaDES. BothDSC and DOSY NMR analyses confirmed the presence of a eutectic point and strong intermolecular interactions,validating the classification of this mixture as a NaDES.Considering Octanoic Acid/1,3-Propanediol Mixture, this system presented a more complex profile. DSCmeasurements indicated the absence of a eutectic-type profile, instead revealing a monotectic-like phase diagramsuggesting a solid-solution nature. Interestingly, DOSY NMR analyses showed that "eutectic" ratio (1:2 mol/mol)simulated by the COSMO-RS, the mixture behaved as a pure substance -a hallmark of eutectic solvents. However,this behavior likely reflects the point of maximum intermolecular interaction rather than a true eutectic point. Theseresults underscore the need for careful experimental validation before assigning the NaDES label to such systems.The results of our multi-technique approach highlight the importance of comprehensive characterization in thestudy of eutectic solvents. By integrating DSC, DOSY NMR, and theoretical modeling, we were able to uncovercritical differences between NaDES and other solvent types. These findings are essential for advancing the designand application of sustainable solvents in green chemistry, ensuring that claims of eco-friendliness and functionalityare backed by robust experimental evidence. In addition, this study supports the adoption of the MiNaCnomenclature as a flexible and inclusive term for mixtures of natural metabolites, especially in cases where thepresence of a eutectic point or glass transition cannot be conclusively demonstrated
Self-Supervised Models of Speech Processing for Haitian Creole
No full textInternational audienceWe develop tailored speech processing models for Haitian Creole positioning it as a high-resource language in terms of SSL models of speech processing. We do so by pretraining monolingual Wav2Vec2-Base, Wav2Vec2-Large and Data2Vec-Audio-Base models from scratch, which are then finetuned on an Automatic Speech Recognition task. We compare the performance of these models with models finetuned from larger multilingual (XLSR-53, XLSR2-300m, MMS-1B) and monolingual French-based models (LeBenchmark 1 to 7K). Our results highlight the effectiveness of pretraining monolingual models from scratch, demonstrating that they can achieve performance comparable to or even surpassing larger models derived from large-scale multilingual pretraining. Our work provides essential resources for robust Haitian Creole speech recognition and offers pretrained models that can be adapted to other French-based Caribbean Creoles, opening new avenues for linguistic research and practical applications in the regio
Mise à disposition des biens loués irrégulière ou cession illicite ?
No full textInternational audienc
Location meublée touristique illicite : pas de condamnation in solidum
No full textInternational audienc