HAL-Paris1
Not a member yet
124252 research outputs found
Sort by
Exploring tourist urbanisation in the South Asian mountains: an introduction
No full textInternational audienc
Special Issue: Exploring tourist urbanisation in the South Asian mountains: an introduction
No full textChercher des données géographiques et se former à leur manipulation: Un panorama partiel des sources de données et de méthodes de traitement de l’information géographique
No full textInternational audienc
Dominique Henry-Gambier, itinéraires d’une paléoanthropologue, du terrain aux interprétations sociales
No full textInternational audienc
Unsupervised spatiotemporal classification of deformation patterns of embryonic tissues matches their fate map
No full textInternational audienceDuring morphogenesis, embryonic tissues display fluid-like behavior with fluctuating strain rates. Digital cell lineages reconstructed from 4D images of developing zebrafish embryos are used to infer representative tissue deformation patterns and their association with developmental events. Finite deformation analysis along cell trajectories and unsupervised machine learning are applied to obtain reduced-order models condensing the collective cell motions, delineating tissue domains with distinct 4D biomechanical behavior. This reduced-order kinematic description is reproducible across specimens and matches fate maps of the zebrafish brain in wild-type and nodal pathway mutants (zoeptz57/tz57), shedding light into the morphogenetic defects causing these mutants’ cyclopia. Furthermore, the inferred kinematic maps also match expression maps of the gene transcription factor goosecoid (gsc). In summary, this work introduces an objective analytical framework to systematically unravel the complex spatiotemporal patterns of embryonic tissue deformations and couple them with cell fate and gene expression maps
COMPUTATIONAL TRUST EVALUATION AND CALIBRATION IN AI-INFUSED SYSTEMS: A SYSTEMATIC LITERATURE REVIEW A PREPRINT
No full textInternational audiencePlacing appropriate trust is critical in the adoption of AI-infused systems, where both over-trust or under-trust can lead to significant harm. Computational trust models provide a trust scalar, enabling a machine to assess both its own operational trust level and the human's trust perception. To synthesize existing computational approaches and their operational insights for well-calibrated trust management in AI-infused systems, we conducted a systematic literature review (SLR) of computational trust models. The key outcome of this review is the derivation of a conceptual framework: the Computational Risk-Sensitive biDirectional Trust Model (CRiDiT). This conceptual model unifies trust evaluation on both machine-side and human-side and conducts a trust calibration mechanism designed to align human perception to the system's self-assessed trustworthiness. CRiDiT integrates an input layer, a computation layer (which computes trust scores and supports trust calibration mechanisms to address over-trust and under-trust), and an output layer in charge of user communication. The CRiDiT model is designed to be a blueprint for future operational guidelines and will be empirically validated through future studies on calibration thresholds and trust cues.</div
Divine Names on the Spot III: Naming and Agency in Ancient Greek and West Semitic Texts
No full textInternational audienceIn the line of the previous volumes of the series “Divine Names on the Spot” devoted to the study of divine names in Greek and Semitic contexts, this third one focuses on the question “who named the gods?” Naming the divine, within the ritual communication or in narratives and discourses about gods and goddesses, involves choices, negotiations or strategies by human agents, in accordance with traditions or in order to activate innovations. Always context-sensitive, the agency of human addressers, narrators, or beneficiaries of the divine powers must be put forward as a main factor of these processes. From the addresses to the gods by kings in Cyprus to the carriage drivers naming Poseidon Helikapanaios in Thessaly, through the carving of divine names on a cup found in Jerusalem, the dozen of contributions gathered here make steps for a long exploration of divine names in the making, and suggest a few directions and orientations for investigating human agency in religious history.Dans la continuité des précédents volumes de la série «Divine Names on the Spot» vouée à l’étude des noms divins en contextes grecs et sémitiques, ce troisième opus est consacré plus précisément à la question «qui nomme les dieux?» Nommer la divinité, dans le cadre de la communication rituelle ou dans les récits et discours sur les dieux et les déesses, implique des choix, des stratégies ou des négociations de la part d’agents humains, en tenant compte de traditions ou de manière à activer des innovations. Toujours dépendante des contextes, l’agentivité des destinateurs humains, des narrateurs ou de quiconque bénéficie de l’action des dieux doit être prise en compte comme l’un des facteurs principaux de ces processus. Depuis les adresses faites aux dieux par les rois de Chypre jusqu’aux conducteurs de chariots nommant Poséidon Helikapanaios en Thessalie, en passant par l’expertise qui a inscrit des noms divins sur une coupe trouvée à Jérusalem, la douzaine de contributions rassemblées ici sont autant d’étapes d’une longue exploration de la fabrique des noms divins, proposant quelques orientations et directions pour l’étude de l’agentivité humaine en histoire des religions
Dynamical analysis of financial stocks network: Improving forecasting using network properties
No full textInternational audienceApplying a network analysis to stock return correlations, we study the dynamical properties of the network and how they correlate with the market return, finding meaningful variables that partially capture the complex dynamical processes of stock interactions and the market structure. We then use the individual properties of stocks within the network along with the global ones, to find correlations with the future returns of individual S&P 500 stocks. Applying these properties as input variables for forecasting, we find a 21 % improvement on the R2score in the prediction of stock returns on long time scales (per year), and 3 % on short time scales (2 days), relative to baseline models without network variables. These findings highlight the potential of integrating network-based variables into stock return prediction models, which could enhance forecasting accuracy and provide a deeper understanding of market dynamics. This approach could be valuable for both investors and researchers seeking to model and predict stock behaviour in complex financial networks
Intelligence artificielle et édition structurée : présentation d’une preuve de concept autour de la revue <em>Gallia – Archéologie des Gaules</em>
No full textInternational audienceStructured editing of scientific journals not only enriches the information itself but also facilitates computational processing, thereby improving the accessibility, interoperability, and reusability of the content. While the journal Gallia is now born-digital in XML-TEI for the publication of its print and online versions, a substantial part of its corpus —corresponding to the oldest issues— exists only as facsimiles made available in PDF format. These scanned documents present a dual challenge: on the one hand, automatic text recognition, often hindered by image quality and the typographical specificities of early publications; on the other hand, aligning this raw text with a rich, standardized structure compliant with the TEI guidelines.In this context, we present a protocol that leverages artificial intelligence to support the conversion of older PDF scans into TEI documents. This semi-automated processing pipeline relies both on advanced OCR techniques and on computer-vision and classification models capable of identifying the editorial structure of articles. Beyond the technical dimension, this protocol contributes to the preservation and enhancement of this scientific heritage. It paves the way for expanded use of the journal’s corpus by facilitating its integration into interoperable digital environments, conducive to cross-referenced research and new forms of analysis.L’édition structurée des revues scientifiques permet non seulement d’enrichir l’information, mais aussi de faciliter son traitement informatique, améliorant ainsi l’accessibilité, l’interopérabilité et la réutilisation des contenus. Si la revue Gallia est aujourd’hui nativement structurée en XML-TEI pour la publication de ses versions imprimée et numérique, une partie importante de son corpus – correspondant aux numéros les plus anciens – n’existe que sous la forme de facsimilés mis à disposition au format PDF. Ces documents scannés posent un double défi : d’une part, la reconnaissance automatique du texte, souvent entravée par la qualité des images et les spécificités typographiques des publications anciennes ; d’autre part, la mise en correspondance de ce texte brut avec une structuration riche et normalisée conforme au standard TEI. Dans ce contexte, nous présentons un protocole mobilisant l’intelligence artificielle afin d’assister la transformation des anciens scans PDF en documents TEI. Cette chaîne de traitement semi-automatique repose à la fois sur des techniques d’OCR avancées et sur des modèles de vision par ordinateur et de classification capables d’identifier la structure éditoriale des articles. Au-delà de l’aspect technique, ce protocole contribue à la pérennisation et à la valorisation de ce patrimoine scientifique. Il ouvre la voie à une exploitation élargie d’un corpus de revue, en facilitant son intégration dans des environnements numériques interopérables, propices aux recherches croisées et à de nouvelles formes d’analyse
Nomad Fertile: nomadisme et pastoralisme au nord de l’ancien croissant fertile
No full texthttps://distam.hypotheses.org/21967La plateforme Nomad Fertile, est destinée à centraliser, partager et valoriser des données de recherches ethno-photographiques sur les pratiques et la culture matérielle des éleveurs pastoraux et nomades de la zone nord de l’ancien croissant fertile (Turquie, Arménie, Azerbaïdjan, Kurdistan irakien). Il s’agit du premier projet déposé sur le portail Distamlango, dédiée aux projets numériques en études aréales développés dans le cadre du consortium Distam