Search

Responsable de la mise en œuvre de l’open data des décisions de justice, la Cour de Cassation développe, depuis 2018, un algorithme d’intelligence artificielle de pseudonymisation automatique de la jurisprudence. À partir d’une enquête ethnographique, cet article rend compte des modalités spécifiques d’articulation du travail des technicien·nes, annotateur·trices et juristes impliqués dans la mise en œuvre de ce projet. L’analyse des activités de catégorisation sous-tendant le fonctionnement de l’IA, alliant opérations théoriques de définition du cadre catégoriel et pratiques empiriques de codage, met en évidence l’importance des choix qui contribuent à donner corps à des outils pourtant considérés comme autonomes, neutres et objectifs. Since 2018, the French Court of Cassation (Cour de Cassation), the institution in charge of giving public access to case law, has been developing an AI algorithm for the automatic anonymization of this data. Based on an ethnographic investigation, this paper documents the articulation of the work conducted by technicians, annotators and legal experts involved in this process. It analyzes the operations of categorization underlying this project, operations which consist both in the abstract determination of the categorial frame, and in the empirical coding practices. It thus shows the importance of the choices that shape algorithmic tools, often considered autonomous, neutral and unbiased.

The main topics of this thesis are modeling and monitoring the health of hybrid systems under uncertainties.Modeling a system is representing the system through a mathematical formalism. A new formalism, the Heteregeneous Petri Nets is introduced in this thesis. This formalism, an extension of the Petri Nets, allows to model and monitor hybrid systems under undertainties. With this formalism, it is possible to represent discrete systems, continuous systems or hybrid systems, for which a new definition is proposed. Contrary to most formalisms, the HtPN formalism allows for a communication between different types of system. Uncertainties, on the model or the observations, can be taken into account to better represent the behavior of the system. Finally, dynamics representing the aging and the wear of the system can be estimated and implemented on the model.For hybrid systems, models based on physical knowledge are often hard to obtain.Artificial intelligence is thus used to learn model from data. This learning is comprised of two steps. The first one is based on a clustering algorithm and outputs the discrete events structure of the model. The second one is based on two different regressions algorithms, support vector and random forest, and learns the continuous dynamics of the system. The choice of which regression algorithm to choose is left to the user, according to his needs.Health monitoring is usually associated to diagnosis, identifying and isolating a problem, usually referred to as a fault, or prognosis, estimating the remaining useful life of a system. An advanced diagnosis methodology, based on the HtPN formalism, is introduced and applied on three different systems. The first system considered is a discrete system. The second one is continuous and the third one is a hybrid system. The hybrid system is a system composed of connected water tanks. In this system, a fault example could be a leakage in one of the tanks. The diagnosis method is done in two steps. The first step is the prediction, which will give hypotheses on the future state of the model. The second step is the correction, in which the different hypotheses given by the first step will either be kept or forgotten, according to the observations received. The diagnosis results obtained are able to take into account uncertainties on the modeling or the observation and gives a belief on different health state hypotheses. Evaluating the degradation of the system can also be done and could be used to apply prognosis methods.Finally, a real case study is introduced. It consists in two communicating photovoltaic panels with some loads: two lamps and a motor. The two panels are modeled using the HtPN formalism. A global model is obtained using the learning method with some collected datasets. Finally, the diagnosis method is applied on data obtained by simulation of the two panels model., Cette thèse concerne la modélisation et la gestion de santé de systèmes hybrides sous incertitude.La modélisation consiste à représenter un système dans un formalisme mathématique. Cette thèse propose un formalisme, les réseaux de Petri Hétérogènes (HtPN), permettant de modéliser et surveiller des systèmes hybrides sous incertitudes. Ce formalisme est une extension des réseaux de Petri classiques et permet de représenter tous les types de systèmes. Il permet donc de représenter des systèmes purement discrets, des systèmes continus ou des systèmes hybrides, dont une définition est proposée, ainsi qu’une communication entre ces systèmes. De plus, il permet de prendre en compte des incertitudes, que ce soit au niveau du modèle ou des observations, afin de représenter au mieux le comportement du système. Le vieillissement et l’usure des systèmes peuvent être estimés et inclus dans le modèle du système sous forme de lois de dégradation.Pour ces systèmes, les modèles issus de connaissances physiques seules sont difficiles à obtenir. L’intelligence artificielle est donc utilisée pour apprendre des modèles à partir de données. Cet apprentissage de modèle est découpé en deux parties : l’apprentissage de la structure à événements discrets du modèle, basé sur un algorithme de clustering, et l’apprentissage des dynamiques d’évolutions à temps continu du système, basé sur deux algorithmes de régression, dont le choix est laissé à l’utilisateur.Par gestion de santé, on entend généralement du diagnostic, qui est de savoir identifier et isoler un problème, appelé une faute, et du pronostic, qui consiste en une estimation de la durée de vie résiduelle du système. Une méthode de diagnostic avancée, basée sur le formalisme des HtPN, est présentée et mise à l’épreuve sur trois systèmes : un système à événements discrets, un système continu et un système hybride. Ce dernier consiste en des réservoirs d’eau connectés entre eux. Une faute peut par exemple correspondre à l’apparition d’une fuite dans un des réservoirs. La méthode de diagnostic se déroule en deux étapes : une étape de prédiction, calculant les différentes hypothèses sur l’état futur du système ; et une étape de correction, dans laquelle les différentes hypothèses sont validées ou réfutées en fonction des observations reçues. Les résultats de diagnostic obtenus permettent de prendre en compte les incertitudes sur la modélisation et sur les observations et de donner des hypothèses de diagnostic pondérées. L’évaluation de la dégradation du système peut également être obtenue, menant à des informations pour le pronostic.Enfin, un cas d’étude réel, composé de deux panneaux photovoltaïques et de diverses charges est présenté. Les deux panneaux ont d'abord été modélisés avec le formalisme des HtPN. Ensuite, un modèle global a été obtenu par apprentissage. Enfin, la méthode de diagnostic avancée est appliquée sur divers cas d'utilisation, montrant la pertinence de la méthode.

Le travail présenté dans cette thèse a pour but de construire un pont entre le contrôle basé sur les données et les logiques de conception de contrôle conventionnelles pour rejeter des perturbations périodiques inconnues complexes agissant sur des systèmes linéaires ou non linéaires. La principale contribution de la thèse est l'approche de contrôle à trois couches qui montre comment combiner le contrôle de type apprentissage (basé sur les données) et le contrôle de type non-apprentissage (conventionnel) au moyen d'un processus d'identification comme outil de connexion entre les deux. Cette approche est expliquée au chapitre 5 de la thèse. Un point particulièrement motivant de l'approche proposée est qu'elle peut fonctionner correctement sous des perturbations périodiques inconnues complexes à la fois lorsque le système est linéaire et non linéaire (les performances en cas linéaire sont beaucoup plus élevées que celles en cas non linéaire pour lesquelles les raisons seront expliquées au chapitre 5 et la conclusion générale). La faisabilité de cette approche est prouvée avec des résultats de simulation. La structure de la thèse est la suivante. Dans le chapitre 1, une introduction est donnée afin de développer les méthodes de contrôle de type apprentissage et non-apprentissage existantes et de décrire le processus de développement de la thèse. Dans le chapitre 2, une étude est consacrée à la bibliographie du contrôle d'apprentissage itératif (ILC) où des concepts de conception importants pour l'ILC ainsi que la principale méthode ILC utilisée tout au long de la thèse (c'est-à-dire la norme optimale ILC (NO-ILC)) sont introduits. Dans le chapitre 3, une application pratique avec un véritable drone d'intérieur est présentée comme une étude d'introduction qui a été réalisée pour tester la faisabilité du NO-ILC et déterminer s'il pouvait ou non être utilisé dans le reste du travail de thèse. Dans le chapitre 4, la principale contribution de la thèse commence à se développer avec la proposition d'une procédure de réglage automatique des contrôleurs de rétroaction linéaire qui reçoit le nom de workflow de réglage de contrôleur basé sur l'apprentissage (LBCT). Cette procédure peut être considérée comme la première tentative de créer une connexion entre le contrôle basé sur les données et le contrôle conventionnel tout au long de la thèse. Il est démontré avec les résultats de simulation d'un problème de rejet de perturbations périodiques inconnu a priori que l'ILC peut être utilisé comme un outil puissant pour simplifier la conception des contrôleurs de rétroaction. L'ensemble de ce processus correspond en fait à la transformation de l'ILC basée sur les données en une loi de rétroaction de manière approximative. Ensuite, dans le chapitre 5, les idées présentées dans le chapitre 4 sont développées plus avant et l'approche qui est mentionnée ci-dessus en tant que contrôle à trois couches est obtenue comme résultat principal de la thèse. La dernière section de la thèse est consacrée à une conclusion qui fournit quelques suggestions sur les études futures dans le sujet donné et elle implique le regard critique de l'auteur sur divers points concernant les applications et les approches menées au cours de la thèse. The work presented in this thesis has the purpose of building a bridge between data-based control and conventional control design logics for rejecting complex unknown periodic disturbances acting on linear or a class of nonlinear systems. The main contribution of the thesis is the triple layer control approach which shows how to combine learning-type (data-based) control and non-learning-type (conventional) control by means of an identification process as connecting tool in between. This approach is explained in Chapter 5 of the thesis. A particularly motivating point of the proposed approach is that it can properly function under complex periodic unknown disturbances both when the system is linear and nonlinear (performance in linear case is much higher from the one in nonlinear case for which the reasons will be explained in Chapter 5 and the general conclusion). The feasibility of this approach is proven with simulation results. The structure of the thesis is as follows. In Chapter 1, an introduction is given in order to elaborate on the existing learning-type and non-learning type control methods and portray the development process of the thesis. In Chapter 2, a study is dedicated to the iterative learning control (ILC) bibliography where important design concepts for ILC as well as the main ILC method used throughout thesis (i.e. the norm optimal ILC (NO-ILC)) are introduced. In Chapter 3, a practical application with a real indoor UAV is shown as an introductory study which was carried out for testing the feasibility of NO-ILC and determining whether or not it could be used in the rest of the thesis work. In Chapter 4, the main contribution of the thesis begins to develop with the proposal of a procedure for automatically tuning the linear feedback controllers which is given the name of learning based controller tuning (LBCT) workflow. This procedure can be seen as the first attempt to create a connection between data-based control and conventional control throughout the thesis. It is demonstrated with the simulation results of an a priori unknown periodic disturbance rejection problem that ILC can be utilised as a powerful tool for simplifying the design of the feedback controllers. This whole process actually corresponds to the transformation of the data-based ILC into a feedback law in an approximate way. Next, in Chapter 5, the ideas presented in Chapter 4 are developed further and the approach that is mentioned above as triple layer control is obtained as the main outcome of the thesis. The final section of the thesis is dedicated to a conclusion that provides some suggestions about future studies in the given topic and it involves the critical view of the author on various points regarding the applications and approaches conducted during the thesis

Nous vivons un temps de changements inédits dans les conceptions de l’apprentissage, de l’École et de l’éducation. Suite à l’affaiblissement de la notion de réforme et à son remplacement par d’autres formes plus pragmatiques d’action politique qui ont conduit à une approche consumériste, nous avons besoin de transformer l’École en pensant l’éducation comme un bien commun. En ce sens, la notion de réforme peut encore être utile. We are living in a time of unprecedented change in how learning, school and education are conceived. Following the weakening of the notion of reform and its replacement by other more pragmatic forms of political action that have led to a consumerist approach, we need to transform the school system by thinking of education as a common good. In this sense, the notion of reform can still be useful. Vivimos un tiempo de cambios inéditos en las concepciones del aprendizaje, de la Escuela y de la educación. A raíz del debilitamiento de la noción de reforma y de su sustitución por otras formas más pragmáticas de acción política que han conducido a una aproximación consumista, necesitamos transformar la Escuela pensando la educación como un bien común. Desde esta perspectiva, la noción de reforma aún puede resultar útil.

Les méthodes d’apprentissage automatique ont été utilisées comme outil de prédiction dans denombreux domaines, mais leur utilisation en santé et sécurité au travail est relativement nouvelle.C’est la raison pour laquelle, il serait intéressant d’utiliser ces méthodes dans la prévision desaccidents du travail, sur les données d’enquêtes européennes sur les conditions de travail. L’objectifde ce travail est de tester les performances des techniques d’apprentissage automatique dans lamodélisation et la prédiction d’accidents du travail. A cette fin, nous utiliserons trois modèles :(les forêts aléatoires) (RF), Support Vector Machine (SVM), modèle logistique. Nous observonsque, la performance des modèles dépend des critères d’évaluation choisis

Rule-based argumentation systems are developed for reasoning about defeasible information. As a major feature, their logical language distinguishes between strict rules (encoding strict information) and defeasible rules (describing general behavior with exceptional cases). They build arguments by chaining such rules, define attacks between them, use a semantics for evaluating the arguments and finally identify the plausible conclusions that follow from the rules. Focusing on the family of inconsistency-based attack relations, this paper presents the first study of the outcomes of such systems under various acceptability semantics, namely naive, stable, semi-stable, preferred, grounded and ideal. It starts by extending the existing list of rationality postulates that any rule-based system should satisfy. Then, it defines the key notion of option of a theory (a theory being a set of facts, a set of strict rules and a set of defeasible rules). For each of the cited semantics, it characterizes the extensions of a rule-based system that satisfies all the postulates in terms of options of the theory under which the system is built. It also fully characterizes the set of plausible conclusions of the system. The results show that designing a rule-based argumentation system requires great care.

Recommender systems aim to support decision-makers by providing decision advice. We review briefly tools of Multi-Criteria Decision Analysis (MCDA), including aggregation operators, that could be the basis for a recommender system. Then we develop a multi-criteria recommender system, STROMa (SysTem of RecOmmendation Multi-criteria), to support decisions by aggregating measures of performance contained in a performance matrix. The system makes inferences about preferences using a partial order on criteria input by the decision-maker. To determine a total ordering of the alternatives, STROMa uses a multi-criteria aggregation operator, the Choquet integral of a fuzzy measure. Thus, recommendations are calculated using partial preferences provided by the decision maker and updated by the system. An integrated web platform is under development.

International audience; Any decision taken by an agent requires some knowledge of its environment. Communication with other agents is a key issue for assessing the overall quality of its own knowledge. This assessment is a challenge itself as the agent may receive information from unknown agents. The aim of this paper is to propose a framework for assessing the reliability of unknown agents based on communication. We assume that information is represented through logical statements and logical inconsistency is the underlying notion of reliability assessment. In our context, assessing consists of ranking the agents and representing reliability through a total preorder.The overall communication set is first evaluated with the help of inconsistency measures. Next, the measures are used for assessing the contribution of each agent to the overall inconsistency of the communication set. After stating the postulates specifying the expected properties of the reliability preorder, we show through a representation theorem how these postulates and the contribution of the agent are interwoven. We also detail how the properties of the inconsistency measures influence the properties of the contribution assessment. Finally we describe how to aggregate different reliability preorders, each of them may be based on different inconsistency measures.; Cet article présente une méthode pour calculer la fiabilité relative de sources d'informations. En se basant sur les degrés d'inconsistance des informations rapportées par les agents, on définit la contribution de chaque agent dans l'inconsistance. Après avoir listé des postulats spécifiant les propriétés attendues d'un pré-ordre de fiabilité, on établit un théorème de représentation qui relie les postulats à cette fonction de contribution. Enfin, on décrit comment agréger différents pré-ordres de fiabilité.

International audience; The paper discusses graded extensions of the cube of opposition, a structure that naturally emerges from the square of opposition in philosophical logic. These extensions of the cube of opposition agree with possibility theory and its four set functions. This extended cube then provides a synthetic and unified view of possibility theory. This is an opportunity to revisit basic notions of possibility theory, in particular regarding the handling of fuzzy events. It turns out that in possibility theory, two extensions of the four basic set functions to fuzzy events exist, which are needed for serving different purposes. The expressions of these extensions involve many-valued conjunction and implication operators that are related either via semi-duality or via residuation.

International audience; Stock market is a complex system composed from heterogeneous traderswith highly non-linear interactions from which emerge a phenomenon of specula-tive bubble. To understand the role of heterogeneous behaviors of traders andinteractions between them in the emergence of bubbles, we propose an agent-basedmodel of double auction market, with asynchronous time management, where tra-ders act asynchronously and take different times to make decisions. The market is populated by heterogeneous traders. In addition to fundamentalist, noise, and technical (chartist) traders, we propose a hybrid trader, which can switch between technical (chartist) and fundamentalist strategies integrating panicking behavior. We find that when market is populated by a majority of hybrid traders, we observequite realistic bubble formation characterized by a boom phase when hybrid tradersswitch to technical behavior, followed by a relatively shorter burst phase whenhybrid traders return to fundamentalist strategy and change to panicked state. The aim is to design agents which act asynchronously, with simple behaviors, but complex enough to produce realistic price dynamics, which provide a basis fordeveloping agents with sophisticated decision-making processes.

We study methods for transforming valued constraint satisfaction problems (VCSPs) to binary VCSPs. First, we show that the standard dual encoding preserves many aspects of the algebraic properties that capture the computational complexity of VCSPs. Second, we extend the reduction of CSPs to binary CSPs described by Bulin et al. [LMCS'15] to VCSPs. This reduction establishes that VCSPs over a fixed valued constraint language are polynomial-time equivalent to Minimum-Cost Homomorphism Problems over a fixed digraph., Subsumes 1507.01776

International audience; In this paper we present a user experience report on a Group Decision Support System. The used system is a Collaborative framework called GRoUp Support (GRUS). The experience consists in three user tests conducted in three different countries. While the locations are different, all three tests were run in the same conditions: same facilitator and tested process. In order to support the end-users. we teach the system in two different ways: a presentation of the system, and a video demonstrating how to use it. The main feedback of this experience is that the teaching step for using Collaborative tools in mandatory. The experience was conducted in the context of decision-making in the agriculture domain.

International audience; The allocation of scarce resources is a complex problem, specially when it comes to budget constraints. Thus, this work aims to propose a multicriteria web-based Decision Support System for resource allocation in the context of higher education organizations, more precisely, public universities that have budget constraints, such as Brazilian federal universities. To do so, the research is divided into three steps: identify the Brazilian general allocation model and the models from each federal university; find similarities between the models; and, divide the models into categories, according to their similarities. Subsequently, a Brazilian federal university was chosen (the Federal University of Mato Grosso do Sul / UFMS) as a parameter to make a numerical application to validate the multicriteria model for resource allocation proposed and, afterward, a web-based DSS was developed. For the MCDM resource allocation model, an additive value function was considered to set the percentage of the total budget that every alternative should receive. The problem was seen as a special case of project portfolio selection problem because its approach is deemed to be appropriate for a resource allocation decision context. Also, the study analyzed the effects of possible scaling issues in additive value functions, when considering resource allocation problems and a sensitivity analysis was performed to analyze the robustness of the model. For the web-based DSS, the analysis was carried out by developing a DSS Database model to store and retrieve data, defining the user’s interface based on his detailed requirement analysis and using a web platform to transform the prototype into a web-based system. The results were achieved. The system provided a clear vision on how the resource allocation procedure works, the entire process became more transparent to the ones that are affected by it, to the decision makers and to the government, enabling them to take safer and reliable decisions, seeking to reduce uncertainties and to maximize their results. The multicriteria web-based DSS presented here could be extended and applied by other federal universities in Brazil or other countries, adapting the alternatives and criteria for each specific internal allocation model and to the DM needs.

The main goal of this research report is to define a generic framework for black-box explanation methods in order to make it easier to compare and classify different approaches. We focus on two components of this framework, called respectively “Sampling” and “Generation”, which are characterized formally and used to build a taxonomy of explanation methods. We also describe precisely how each method can be expressed in the framework.; Dans ce rapport de recherche, nous proposons un cadre générique pour décrire les méthodes d’explication fonctionnant en mode “boîte noire”, l’objectif étant de faciliter la comparaison et la classification de ces méthodes. Nous définissons formellement deux composantes principales, appelées respectivement l’“Echantillonnage” et la “Génération”, qui sont ensuite utilisées pour construire une taxonomie des méthodes d’explication. Nous décrivons aussi précisément la manière dont les méthodes de la littérature s’expriment dans ce cadre.

The main goal of this research report is to define a generic framework for black-box explanation methods in order to make it easier to compare and classify different approaches. We focus on two components of this framework, called respectively “Sampling” and “Generation”, which are characterized formally and used to build a taxonomy of explanation methods. We also describe precisely how each method can be expressed in the framework.; Dans ce rapport de recherche, nous proposons un cadre générique pour décrire les méthodes d’explication fonctionnant en mode “boîte noire”, l’objectif étant de faciliter la comparaison et la classification de ces méthodes. Nous définissons formellement deux composantes principales, appelées respectivement l’“Echantillonnage” et la “Génération”, qui sont ensuite utilisées pour construire une taxonomie des méthodes d’explication. Nous décrivons aussi précisément la manière dont les méthodes de la littérature s’expriment dans ce cadre.

Recently, educational videos have become important parts of e-learning systems which have in turn become widely used due to their flexibility. These videos should be of high quality since higher production values lead to superior learning outcomes. However, creating high-quality video is a difficult task for teachers since it needs technical knowledge that includes video recording and timeline usage. Hence, creating educational video production software, that is at the same time easy-to-use and able to produce high-quality educational videos, is very advantageous. In this paper, we developed protocols for an easy-to-use piece of software that enables teachers who have little technological background to produce their own educational videos autonomously. In fact, our contribution is to reduce the complexity of the whole video production process by introducing a preparation step based on micro-teaching and upstream specification. An evaluation of the software with six teachers is performed. This evaluation, based on think-aloud protocol and quantitative measurements, showed that the introduction of the preparation step allowed the participant teachers to produce high-quality educational videos in less than three hours.

International audience; Datacleaninganddatapreparationhavebeenlong-standingchallenges in data science to avoid incorrect results, biases, and misleading conclusions ob- tained from “dirty” data. For a given dataset and data analytics task, a plethora of data preprocessing techniques and alternative data cleaning strategies are avail- able, but they may lead to dramatically different outputs with unequal result quality performances. For adequate data preparation, the users generally do not know how to start with or which methods to use. Most current work can be classified into two categories: 1) they propose new data cleaning algorithms specific to certain types of data anomalies usually considered in isolation and without a “pipeline vision” of the entire data preprocessing strategy; 2) they develop automated machine learning approaches (AutoML) that can optimize the hyper- parameters of a considered ML model with a list of by-default preprocessing methods. We argue that more efforts should be devoted to proposing a principled and adaptive data preparation approach to help and learn from the user for selecting the optimal sequence of data preparation tasks to obtain the best quality performance of the final result. In this paper, we extend Learn2Clean, a method based on Q-Learning, a model-free reinforcement learning technique that selects, for a given data set, a given ML model, and a pre-selected quality performance metric, the optimal sequence of tasks for preprocessing the data such that the quality metric is maximized. We will discuss some new results of Learn2Clean for semi-automating data preparation with “the human in the loop” using active reward learning and Q-learning.

With the evolution of technologies and robotics, the possibilities offered by 3D acquisition systems have increased. Nowadays, these systems are used in different domains as for autonomous vehicles, rescue robots, cultural heritage, for example. These application fields often require to perform object recognition from acquired data. Therefore, various methodologies have been investigated to automatically process 3D point cloud data in order to detect contained objects. The best methodologiesdepend on the context, that means they are specific to the data to be processed and the objects to be recognized. They produce efficient recognition, which is essential whatever the application field. However, adapting methodologies to a particular application field or use case limits the flexibility to extend the use of a method to other fields. These observations highlight the importance of developing object recognition methodologies specific to a detection context, but also the limitation of existing methods to preserve their capacity within changing detection contexts. An excellent example of a high degree of flexibility to changing contexts is human intelligence and human’s ability to design ad hoc methodologies. Humans can analyze the context according to their knowledge and combine different characteristics or strategies according to the objective to be achieved. It would, therefore, be helpful for Computer Vision tools to integrate elements of artificial intelligence, allowing to adapt to the context of an application fields and to guide the detection process in this respect. This Ph.D. thesis presents a knowledge-based approach for object recognition that can be used whatever the application field. Its architecture is based on semantic technologies to allow a knowledge management module to guide the objects detection process through a step by step procedure performing the selection, parameterization, and execution of algorithms. The detection process is performed thanks to an artificial intelligence approach that uses explicit knowledge to design a context-dependent object recognition solution. Its strength is its adaptability to the context, but also its capability to analyze and understand a scene and contained objects and the specificities of the data to be processed. This understanding capability is realized through a self-learning process able to define and validate hypotheses concerning the context, also enabling to enrich the knowledge base and to improve the objects recognition process. The efficiency of this adaptation capability will be demonstrated in four use cases from different application fields. The first use case is an indoor of a building. It is used for a monitoring purpose. The second use case is located in the field of Archaeology represented by ancient ruins containing a terrace house with a watermill. The third use case is an outdoor representing a part of the city of Freiburg in Germany. It is used for an industrial purpose. Finally, the last use case is an indoor acquired by Microsoft’s Kinect. It is used for a robotic purpose., Avec l’évolution des technologies et de la robotique, les possibilités offertes par les systèmes d’acquisition 3D ont augmenté. Aujourd’hui, ces systèmes sont utilisés dans différents domaines comme par exemple pour les véhicules autonomes,les robots de sauvetage, le patrimoine culturel. Ces champs d’application nécessitent souvent la reconnaissance d’objets à partir de données acquises. C’est pourquoi diverses méthodologies ont été étudiées pour traiter automatiquementles données 3D des nuages de points afin de détecter les objets contenus. Les meilleures méthodologies dépendent du contexte, c’est-à-dire qu’elles sont spécifiques aux données à traiter et aux objets à reconnaître. Elles produisent une reconnaissance performante, ce qui est essentiel quel que soit le domaine d’application. Toutefois, l’adaptation des méthodologies à un domaine d’application ou à un cas d’utilisation particulier limite la possibilité d’étendre l’utilisation d’une méthode à d’autres domaines. Ces observations soulignent l’importance de développer des méthodologies de reconnaissance d’objets spécifiques à un contexte de détection, mais aussi la limitation des méthodes existantes pour préserver leur capacité dans des contextes de détection changeants. Un excellent exemple d’un degré élevé de flexibilité face à l’évolution des contextes est l’intelligence humaine et la capacité de l’homme à concevoir des méthodologies ad hoc. L’homme peut analyser le contexte en fonction de ses connaissances et combiner différentes caractéristiques ou stratégies en fonction de l’objectif à atteindre. Il serait donc utile que les outils de vision par ordinateur intègrent des éléments d’intelligence artificielle permettant de s’adapter au contexte d’un domaine d’application et de guider le processus de détection à cet égard. Cette thèse de doctorat présente une approche de la reconnaissance d’objets basée sur la connaissance qui peut être utilisée dans tous les domaines d’application. Son architecture est basée sur des technologies sémantiques pour permettre à un module de gestion des connaissances de guider le processus de détection d’objets à travers une procédure étape par étape effectuant la sélection, le paramétrage et l’exécution des algorithmes. Le processus de détection est réalisé grâce à une approche d’intelligence artificielle qui utilise des connaissances explicites pour concevoir une solution de reconnaissance d’objets en fonction du contexte. Sa force réside dans son adaptabilité au contexte, mais aussi dans sa capacité d’analyse et de compréhension d’une scène et d’objets contenus ainsi que dans les spécificités des données à traiter. Cette capacité de compréhension est réalisée par un processus d’auto-apprentissage capable de définir et de valider des hypothèses concernant le contexte, permettant ainsi d’enrichir la base de connaissances et d’améliorer le processus de reconnaissance des objets. L’efficacité de cette capacité d’adaptation sera démontrée dans quatre cas d’utilisation de différents domaines d’application. Le premier cas d’utilisation est l’intérieur d’un bâtiment. Il est utilisé à des fins de surveillance. Le second cas d’utilisation se situe dans le domaine de l’archéologie représenté par des ruines anciennes contenant une maison en terrasse avec un moulin à eau. Le troisième cas d’utilisation est un extérieur représentant une partie de la ville de Fribourg en Allemagne. Il est utilisé à des fins industrielles. Enfin, le dernier cas d’utilisation est un intérieur acquis par Kinect de Microsoft. Il est utilisé à des fins robotiques.

International audience; In this paper we study the link between a bipolar decision structure called BLF (bipolar leveled framework) and the qualitative decision theory based on possibility theory. A BLF defines the set of possible decision principles that may be used in order to evaluate the admissibility of a given candidate. A decision principle is a rule that relates some observations about the candidate to a given goal that the selection of this candidate may achieve or miss. The decision principles are ordered according to the importance of the goal they support. Oppositions to decision principles are also described in the BLF under the form of observations that contradict the realization of the decision principle. In order to show that this rich and visual framework is well founded we show how the notions defined in the BLF can be translated in terms of qualitative decision theory.

© 2015 Published by Elsevier B.V. Sugeno integrals and their particular cases such as weighted minimum and maximum have been used in multiple-criteria aggregation when the evaluation scale is qualitative. This paper proposes two new variants of weighted minimum and maximum, where the criteria weights play the role of tolerance thresholds. These variants require the use of a residuated structure, equipped with an involutive negation. We propose residuated counterparts of Sugeno integrals, where the weights bear on subsets of criteria, and we study their properties, showing they are analogous to Sugeno integrals to a large extent. Finally we propose dual aggregation operations, we call desintegrals, where an item is evaluated in terms of its defects rather than in terms of its positive features. Desintegrals are maximal when no defects at all are present, while integrals are maximal when all merits are sufficiently present. Qualitative integrals and desintegrals suggest a possible approach to bipolar evaluation processes where items are judged both in terms of merits and defects that are not independent of one another.

New Approaches in Cognitive Robotics . This issue presents a set of contributions showing recent evolutions of researches within the so-called “ Cognitive Robotics” field. This term refers to any scientific research aiming at having a robot realize a task which is considered, when performed by humans, to involve cognitive functions. The latters can be (without being exhaustive) learning, social interaction, perception, motricity, spatial cognition, reasoning, language, or awareness. The goal of this issue is to show how the specific objectives and methods of this research field have historically and progressively stepped away from other work in Robotics and Artificial Intelligence, in order to get closer to other disciplines within the field of Cognitive Sciences. As a corollary, the issue aims at making explicit some identified bridges that could enable these various disciplines to cross-fertilize. In particular, one of the objectives of this issue is to illustrate how robotics experimentation can serve as platforms to test hypotheses from other disciplines of Cognitive Sciences, and thus contribute to the study of biological cognition., Ce volume présente un ensemble de contributions montrant les évolutions récentes des recherches du domaine dit de «Robotique Cognitive ». Cette dénomination vaut dès lors que l’on cherche à faire réaliser au robot des tâches qui semblent nécessiter chez l’homme la mise en oeuvre de fonctions cognitives telles que (sans être exhaustifs) l’apprentissage, l’interaction sociale, la perception, la motricité, la cognition spatiale, le raisonnement, le langage ou encore la conscience. Le but du volume est de montrer en quoi les objectifs et méthodes particuliers de ce domaine se sont historiquement distingués d’autres travaux en Robotique ou en Intelligence Artificielle, pour se rapprocher des autres disciplines des Sciences Cognitives. En corollaire, le volume vise à expliciter certains des ponts possibles qui peuvent permettre à ces disciplines de se féconder mutuellement. En particulier, un des objectifs de ce volume est d’illustrer en quoi l’expérimentation robotique peut servir de plateforme de test d’hypothèses d’autres disciplines des Sciences Cognitives, et ainsi contribuer à l’étude de la cognition biologique., Réguigne-Khamassi Mehdi, Doncieux Stéphane. Nouvelles approches en Robotique Cognitive. In: Intellectica. Revue de l'Association pour la Recherche Cognitive, n°65, 2016/1. Nouvelles approches en Robotique Cognitive. pp. 7-25.

National audience; Le plan numérique pour l’éducation lancé en mai 2015 vise à introduire les technologies numériques au sein des établissements scolaires. Dans cette perspective, les élèves sont progressivement équipés d’outils informatiques, les enseignants sont formés aux usages de ces technologies pour l’enseignement (p. ex. tableaux interactifs, environnement numérique de travail, etc.) et des ressources telles que les manuels scolaires numériques sont mises à disposition des enseignants et des élèves. Or, depuis le plan Handiscol (1999), puis avec la loi du 11 février 2005 pour l'égalité des droits et des chances, la participation et la citoyenneté des personnes handicapées et la loi d'orientation et de programmation pour la refondation de l'école de la République (2013), les politiques relatives au handicap promeuvent l’inclusion des élèves en situation de handicap en établissements scolaires ordinaires. C’est ainsi qu’en 2015, 80% des 350 300 enfants ou adolescents handicapés ont été scolarisés en établissements ordinaires. Si les politiques d’inclusion prévoient diverses mesures afin de garantir « l’égalité des droits et des chances », qu’en est-il de l’accessibilité des ressources pédagogiques numériques pour ces élèves à besoins éducatifs particuliers ? Une étude exploratoire combinant des entretiens avec divers acteurs de la situation (enseignants spécialisés, transcripteurs braille, élèves déficients visuels scolarisés en classe ordinaire) et des observations en classe d’élèves présentant une déficience visuelle a été menée dans le but 1) de décrire la situation existante en terme d’accès aux contenus pédagogiques pour les élèves présentant une déficience visuelle (quels contenus sont accédés et comment), 2) d’identifier des catégories de difficultés rencontrées par les élèves pour accéder et traiter les ressources pédagogiques numériques et leurs conséquences pour l’apprentissage.

International audience; Finding a solution to a constraint satisfaction problem (CSP) is known to be an NP-hard task. Considerable effort has been spent on identifying tractable classes of CSP,in other words, classes of constraint satisfaction problems for which there are polynomial time recognition and resolution algorithms. In this article, we present a relational tractable class of binary CSP. Our key contribution is a new ternary operation that we name mjx. We first characterize mjx-closed relations which leads to an optimal algorithm to recognize such relations. To reduce space and time complexity, we define a new storage technique for these relations which reduces the complexity of establishing a form of strong directional path consistency, the consistency level that solves all instances of the proposed class (and, indeed, of all relational classes closed under a majority polymorphism).

International audience; Dung’s abstract argumentation provides us with a general framework to deal with argumentation, non-monotonic reasoning and logic programming. For the extension-based semantics, one of the basic principles is I-maximality which is in particular related with the notion of skeptical justification. Another one is directionality which can be employed for the study of dynamics of argumentation. In this paper, we introduce two new extension-based semantics into Dung’s abstract argumentation, called grounded-like semantics and initial semantics which satisfy the I-maximality and directionality principles. The initial semantics has many good properties and can be expected to play a central role in studying other extension-based semantics, such as admissible, complete and preferred semantics.

Artificial Intelligence (AI) approaches to problem-solving and decision-making are becoming more and more complex, leading to a decrease in the understandability of solutions. The European Union's new General Data Protection Regulation tries to tackle this problem by stipulating a "right to explanation" for decisions made by AI systems. One of the AI paradigms that may be affected by this new regulation is Answer Set Programming (ASP). Thanks to the emergence of efficient solvers, ASP has recently been used for problem-solving in a variety of domains, including medicine, cryptography, and biology. To ensure the successful application of ASP as a problem-solving paradigm in the future, explanations of ASP solutions are crucial. In this survey, we give an overview of approaches that provide an answer to the question of why an answer set is a solution to a given problem, notably off-line justifications, causal graphs, argumentative explanations and why-not provenance, and highlight their similarities and differences. Moreover, we review methods explaining why a set of literals is not an answer set or why no solution exists at all., Comment: Under consideration in Theory and Practice of Logic Programming (TPLP)

International audience; This paper investigates an issue at the interface between language and ontology. It is argued that the phenomenon of 'inherent polysemy' observed in lexical semantics for nouns such as book or country actually is a deeper phenomenon grounded on specific ontological relations involving the entities referred to. It is shown that this phenomenon emerges not only in language but also in most available ontologies. Beyond the 'dot types' used in some linguistic theories to account for logical polysemy, it is proposed to introduce 'complex categories' in ontologies in order to solve incoherence and inconsistency issues appearing when this phenomenon is not acknowledged and to characterize complex categories on the basis of formal ontology relations.

© 2015, Springer Science+Business Media Dordrecht. This paper clarifies the connection between multiple criteria decision-making and decision under uncertainty in a qualitative setting relying on a finite value scale. While their mathematical formulations are very similar, the underlying assumptions differ and the latter problem turns out to be a special case of the former. Sugeno integrals are very general aggregation operations that can represent preference relations between uncertain acts or between multifactorial alternatives where attributes share the same totally ordered domain. This paper proposes a generalized form of the Sugeno integral that can cope with attributes having distinct domains via the use of qualitative utility functions. It is shown that in the case of decision under uncertainty, this model corresponds to state-dependent preferences on consequences of acts. Axiomatizations of the corresponding preference functionals are proposed in the cases where uncertainty is represented by possibility measures, by necessity measures, and by general order-preserving set-functions, respectively. This is achieved by weakening previously proposed axiom systems for Sugeno integrals.

International audience; This paper explores flows and stocks, at the scale of the European Union, of certain rare earth elements (REEs; Pr, Nd, Eu, Tb, Dy and Y) which are associated with products that are important for the decarbonisation of the energy sector and that also have strong recycling potential. Material flow analyses were performed considering the various steps along the value chain (separation of rare earth oxides, manufacture of products, etc.) and including the lithosphere as a potential stock (potential geological resources). Results provide estimates of flows of rare earths into use, in-use stocks and waste streams. Flows into use of, e.g., Tb in fluorescent lamp phosphors, Nd and Dy in permanent magnets and Nd in battery applications were estimated, for selected reference year 2010, as 35, 1230, 230 and 120 tons respectively. The proposed Sankey diagrams illustrate the strong imbalance of flows of permanent magnet REEs along the value chain, with Europe relying largely on the import of finished products (magnets and applications). It is estimated that around 2020, the amounts of Tb in fluorescent lamps and Nd in permanent magnets recycled each year in Europe, could be on the order of 10 tons for Tb and between 170 and 230 tons for Nd.

© 2015, Springer International Publishing Switzerland. The paper shows that a cube of opposition, a structure that generalizes the square of opposition invented in ancient logic, can be generated from the composition of a binary relation with a subset, by the effect of set complementation on the subset, on the relation, or on the result of the composition. Since the composition of relations is encountered in many areas, the structure of opposition exhibited by the cube of opposition has a universal flavor. In particular, it applies to information processing-oriented settings such as rough set theory, possibility theory, or formal concept analysis. We then discuss how this structure extends to a fuzzy relation and a fuzzy subset, and the graded cube of opposition thus obtained provides an organized view of the different existing compositions of fuzzy relations. The paper concludes by pointing out areas of research where the cube of opposition, or its graded version are of interest.

International audience; We devote this paper to the axiomatization/completeness of PRSPDL0 - a variant of iteration-free PDL with parallel composition. Our results are based on the following: although the program operation of parallel composition is not modally definable in the ordinary language of PDL, it becomes definable in a modal language strengthened by the introduction of propositional quantifiers. Instead of using axioms to define the program operation of parallel composition in the language of PDL enlarged with propositional quantifiers, we add an unorthodox rule of proof that makes the canonical model standard for the program operation of parallel composition and we use large programs for the proof of the Truth Lemma.

Logistic provider selection is Multi-Criteria Decision-Making problem.Selection of the best alternative depends on opinion of the group stakeholders.Criteria values are determined in linguistic terms and real use of DSS presented.Basis of the developed DSS is the second most popular MCDM method TOPSIS-F.The investigation carried out under a European Commission project. PurposeThe paper proposes a decision support system for selecting logistics providers based on the quality function deployment (QFD) and the technique for order preference by the similarity to ideal solution (TOPSIS) for agricultural supply chain in France. The research provides a platform for group decision making to facilitate decision process and check the consistency of the outcomes. MethodologyThe proposed model looks at the decision problem from two points of view considering both technical and customer perspectives. The main customer criteria are confidence in a safe and durable product, emission of pollutants and hazardous materials, social responsibility, etc. The main technical factors are financial stability, quality, delivery condition, services, etc. based on the literature review. The second stage in the adopted methodology is the combination of quality function deployment and the technique for order preference by similarity to ideal solution to effectively analyze the decision problem. In final section we structure a group decision system called GRoUp System (GRUS) which has been developed by Institut de Recherche en Informatique de Toulouse (IRIT) in the Toulouse University. ResultsThis paper designs a group decision making system to interface decision makers and customer values in order to aid agricultural partners and investors in the selection of third party logistic providers. Moreover, we have figured out a decision support system under fuzzy linguistic variables is able to assist agricultural parties in uncertain situations. This integrated and efficient decision support system enhances quality and reliability of the decision making. Novelty/OriginalityThe novelty of this paper is reflected by several items. The integration of group multi-criteria decision tools enables decision makers to obtain a comprehensive understanding of customer needs and technical requirements of the logistic process. In addition, this investigation is carried out under a European commission project called Risk and Uncertain Conditions for Agriculture Production Systems (RUC-APS) which models risk reduction and elimination from the agricultural supply chain. Ultimately, we have implemented the decision support tool to select the best logistic provider among France logistics and transportation companies.

An elementary mathematical example proves, thanks to the Routh-Hurwitz criterion, a result that is intriguing with respect to today's practical understanding of model-free control, i.e., an "intelligent" proportional controller (iP) may turn to be more difficult to tune than an intelligent proportional-derivative one (iPD). The vast superiority of iPDs when compared to classic PIDs is shown via computer simulations. The introduction as well as the conclusion analyse model-free control in the light of recent advances., Comment: in French, https://www.openscience.fr/Un-resultat-intrigant-en-commande-sans-modele

International audience; An elementary mathematical example proves, thanks to the Routh-Hurwitz criterion, a result that is intriguing with respect to today's practical understanding of model-free control, i.e., an " intelligent " proportional controller (iP) may turn to be more difficult to tune than an intelligent proportional-derivative one (iPD). The vast superiority of iPDs when compared to classic PIDs is shown via computer simulations. The introduction as well as the conclusion analyse model-free control in the light of recent advances.; Un exemple mathématique élémentaire prouve, grâce au critère de Routh-Hurwitz, un résultat à l'encontre de la pratique actuelle en commande sans modèle : il peut y avoir plus de difficultés à régler un correcteur proportionnel « intelligent » (iP) qu'un proportionnel-dérivé intelligent (iPD). Les simulations numériques de l'iPD et d'un PID classique tournent largement en faveur du premier. Introduction et conclusion analysent la commande sans modèle à la lumière des avancées actuelles.

International audience; We study the use of a bipolar decision structure called BLF (bipolar leveled framework) in the context of collective decision making where the vote consists in giving factual information about a candidate which the group should accept or reject. A BLF defines the set of possible decision principles that may be used in order to evaluate the admissibility of a given candidate. A decision principle is a rule that relates some observations about the candidate to a given goal that the selection of this candidate may achieve or miss. The decision principles are ordered accordingly to the importance of the goal they support. Oppositions to decision principles are also described in the BLF under the form of observations that contradict the realization of the decision principles. We show how the use of a common BLF may reduce the impact of manipulation strategies in the context of group decision making.

Far from being a game, betting exchange world is a real business where there is no place for something else than technical capabilities to learn and forecast. After some successful experiments, a study and research phase has been started in order to discover betting exchange mechanisms. The most important requirement was to build Qwin as a fully automated system, with no human interaction. Goal was simply to conceive a forecast and trading system based on player performance and real-time financial analysis. During 2 years, research, analyse, conception and development activities allowed to master how betting exchange works, using the Betfair API. To get the best possible predictions, a long time is needed in collecting data. Time ran out, and the system had to be stopped by lack of consistent data of financial resources. Initial set of information was enough to start projet but not to be confident in production despite encouraging results. 2 years collecting data was expected, and only 10 months were available.But no doubt the system will restart because it’s fully operational and it has proved it provides high performance forecasts. The thesis is describing in depth all the project phases, how betting exchange works, how the conception and development has been realised, based on today computing technologies. It provides a detailed appreciation on what it’s important to remember regarding methods for predicting : what, how and when.; Loin d’être un jeu, le monde des paris sportifs est une véritable entreprise financière où les émotions laissent place à la technique, à l’apprentissage et à la prédiction. Après quelques expérimentations réussies, un travail d’étude et de recherche a été entrepris afin de découvrir les mécanismes des places de marchés sportifs. Les exigences initiales du projet précisaient clairement que le système Qwin devait être autonome et ne nécessiter aucune intervention humaine. L’objectif était simple, développer un système de prédiction et de placement de paris sportifs, basé sur une analyse des performances des équipes, et des évolutions financières des marchés en temps-réel. Durant 2 années, un travail de recherche, d’analyse, de conception, de développement et d’exploitation a été réalisé et a permis d’acquérir la maîtrise du fonctionnement des marchés sportifs ainsi que de mettre au point un ensemble de programmes pleinement fonctionnels, qualifiés, testés et performants autour de l’API de l’Exchange Betfair. Il a néanmoins fallu, par manque de fonds, mettre en standby le système, car, pour disposer des meilleures prédictions, une profondeur de base de données de plusieurs années est indispensable au simulateur. Le temps a manqué. Les bases utilisées au démarrage du projet ne contenaient pas assez d’information. Une phase de collecte supplémentaire de 2 ans aurait été nécessaire pour assurer une fiabilité des prédictions et une remise en route des applications. Malgré des résultats plus qu’encourageants, seulement 10 mois de données complètes ont pu être collectées. Le projet Qwin redémarrera très certainement. Un long travail ainsi qu’un investissement conséquent ont été nécessaire à la mise en place de ce projet. Ce mémoire décrit l’ensemble des phases du projet, détaille la partie développement et le fonctionnement des marchés sportifs et donne une appréciation sur ce qu’il faut retenir des méthodes et techniques d’apprentissage et de prédiction.

International audience; Extracting hypernym relations from text is one of the key steps in the automated construction and enrichment of semantic resources. The state of the art offers a large varierty of methods (linguistic, statistical, learning based, hybrid). This variety could be an answer to the need to process each corpus or text fragment according to its specificities (e.g. domain granularity, nature, language, or target semantic resource). Moreover, hypernym relation may take different linguistic forms. The aim of this paper is to study the behaviour of a supervised learning approach to extract hypernym relations whatever the way they are expressed, and to evaluate its ability to capture regularities from the corpus, without human intervention. We apply a distant supervised learning algorithm on a sub-set of Wikipedia in French made of disambiguation pages where we manually annotated hypernym relations. The learned model obtained a F-measure of 0.67, outperforming lexico-syntactic pattern matching used as baseline.

We introduce a first-order extension of dynamic logic (FO-DL), suitable to represent and reason about the behaviour of Data-aware Systems (DaS), which are systems whose data content is explicitly exhibited in the system’s description. We illustrate the expressivity of the formal framework by modelling English auctions as DaS, and by specifying relevant properties in FO-DL. Most importantly, we develop an abstraction-based verification procedure, thus proving that the model checking problem for DaS against FO-DL is actually decidable, provided some mild assumptions on the interpretationdomain.

In the typical framework for boolean games (BG) each player can change the truth value of some propositional atoms, while attempting to make her goal true. In standard BG goals are propositional formulas, whereas in iterated BG goals are formulas of Linear Temporal Logic. Both notions of BG are characterised by the fact that agents have exclusive control over their set of atoms, meaning that no two agents can control the same atom. In the present contribution we drop the exclusivity assumption and explore structures where an atom can be controlled by multiple agents. We introduce Concurrent Game Structures with Shared Propositional Control (CGS-SPC) and show that they ac- count for several classes of repeated games, including iterated boolean games, influence games, and aggregation games. Our main result shows that, as far as verification is concerned, CGS-SPC can be reduced to concurrent game structures with exclusive control. This result provides a polynomial reduction for the model checking problem of specifications in Alternating-time Temporal Logic on CGS-SPC., In Proceedings TARK 2017, arXiv:1707.08250

International audience; In this paper, we deal with the problem of tactical capacitated production planning with the demand under uncertainty modelled by closed intervals. We propose a single-item with backordering model under small uncertainty in the cumulative demand for the Master Production Scheduling (MPS) problem with different rules, namely the Lot For Lot rule and the Periodic Order Quantity rule. Then we study a general multilevel, multi-item, multi-resource model with backordering and the external demand on components for the Material Requirement Planning (MRP) problem under uncertainty in the cumulative demand. In order to choose robust production plans for the above problems that hedge against uncertainty, we adopt the well-known minmax criterion. We propose polynomial methods for evaluating the impact of uncertainty on a given production plan in terms of its cost and for computing optimal robust production plans for both problems (MPS/MRP) under the assumed interval uncertainty representation. We show in this way that the robust problems (MPS/MRP) under this uncertainty representation are not much computationally harder than their deterministic counterparts.

Le web de données liées se compose d'entrepôts de données décrits par des ontologies. Ces on- tologies hétérogènes ont différents niveaux de disparité (différences terminologiques, de conceptualisation, de modélisation). Les alignements simples font correspondre une entité de l'ontologie source à une entité de l'on- tologie cible. Les alignements complexes, complètent les alignements simples en exprimant plus finement les différents types de disparités. Des approches permettant de détecter des correspondances complexes entre on- tologies émergent et il n'existe pas encore de jeu de données exhaustif pour les évaluer. Cet article présente un jeu de données d'alignements complexes entre deux paires d'ontologies du domaine de l'organisation de conférences, ainsi qu'une évaluation d'approches d'alignement permettant d'obtenir de telles correspondances.

International audience; Scientific texts represent a rich source of unstructured knowledge. Extracting this knowledge in a supervised manner can become highly expensive in time and human resources. Moreover supervised models are domain- and language-dependent which make them hard to maintain and extend. Hence unsupervised methods have received a lot of attention from researchers in the fields of information extraction and data mining. In this paper, we present our experiments with self-organizing maps (SOMs) for the task of open relation extraction. We combine contextual features of different level (lemmas and parts-of-speech) to help the algorithm to automatically discover lexical and morphological patterns in the corpus. The evaluation results show that our model yields a better performance than the widely used K-means clustering algorithm with the same feature set.

International audience; Today, organizations have become more and more complex which has complicated the decision making process. However, with the technological advances and the evolution of collaboration technologies, industrial companies wish to benefit from it, in order to accelerate the interventions of the specialists of maintenance on the sites and to reduce the time inactivity of their equipment. As a result, this makes it possible to envisage keeping them at a distance without necessarily having locally qualified staff. Indeed, these situations require the rapid and effective intervention of experts, who are not basically available. Furthermore, in this work we propose a new service oriented approach which is essentially based on mobiles agents.

In this paper, we propose a single-agent logic of goal-directed knowing how extending the standard epistemic logic of knowing that with a new knowing how operator. The semantics of the new operator is based on the idea that knowing how to achieve $\phi$ means that there exists a (uniform) strategy such that the agent knows that it can make sure $\phi$. We give an intuitive axiomatization of our logic and prove the soundness, completeness, and decidability of the logic. The crucial axioms relating knowing that and knowing how illustrate our understanding of knowing how in this setting. This logic can be used in representing both knowledge-that and knowledge-how., Comment: an earlier version of the paper to appear in IJCAI 2017

International audience; The mental attitudes of belief, desire, and intention play a central role in the design and implementation of autonomous agents. In 1987 Bratman proposed their integration into a belief-desire-intention (BDI) theory that was seminal in AI. Since then numerous approaches were built on the BDI paradigm, both practical (BDI architectures and BDI agents) and formal (BDI logics). The logical approaches that were most influential are due to Cohen&Levesque and to Rao&Georgeff. However, three fundamental problems remain up to now. First, the practical and the formal approaches evolved separately and neither fertilized the other. Second, only few formal approaches addressed some important issues such as the revision of intentions or the fundamentally paraconsistent nature of desires, and it seems fair to say that there is currently no consensual logical account of intentions. Finally, only few publications study the interaction between intention and other concepts that are naturally connected to intention, such as actions, planning, and the revision of beliefs and intentions. Our paper summarizes the state of the art, discusses the main open problems, and sketches how they can be addressed. We argue in particular that research on intention should be better connected to fields such as reasoning about actions, automated planning, and belief revision and update.

International audience; Figurative language detection has gained relevance recently, due to its importance for efficient sentiment analysis. For the first time, the Défi Fouille de Texte (DEFT) shared task aims to analyse the impact of figuratif language (focusing in particular on irony, sarcasm and humor) on sentiment analysis of French tweets. Three tasks with an increasing level of complexity have been proposed to participants : (T1) polarity analysis of non figurative tweets, (T2) figurative language detection, and (T3) polarity analysis of non figurative and figurative tweets. Twelve teams participated in the competition. Best results in terms of macro f-score are 0.650 for (T1), 0.783 for (T2) and 0.594 for (T3). These results show that the presence of figurative devices make sentiment analysis of tweets much more complex.; La détection automatique du langage figuratif dans les réseaux sociaux est un sujet de recherche extrêmement actif principalement en raison de son importance pour améliorer les performances des systèmes d'analyse d'opinions. Pour la première fois, l'édition 2017 du Défi Fouille de Texte (DEFT) s'intéresse à l'influence du langage figuratif (en particulier l'ironie, le sarcasme et l'humour) dans l'analyse d'opinions à partir de tweets en français. Trois tâches de niveaux de complexité croissants ont été proposées aux participants : (T1) déterminer la polarité globale des tweets non figuratifs, (T2) déterminer si un tweet contient ou non du langage figuratif, et (T3) déterminer la polarité globale des tweets figuratifs et non figuratifs. Douze équipes ont participé à ce défi. Les meilleurs résultats, en macro f-mesure, sont de 0,650 pour (T1), 0,783 pour (T2) et 0,594 pour (T3). Ces résultats montrent clairement que l'usage du langage figuratif complique considérablement l'analyse d'opinions.

International audience; Boolean Region Connection Calculus is a formalism for reasoning about the topological relations between regions. In this paper, we provide computability results about unifiability in Boolean Region Connection Calculus and prove that elementary unification is finitary

International audience; Configuration of a system for decision making problem which facilitates decision process and enables users to get higher quality advantages is a deal always in academic communities. This paper deals with proposing and constructing decision making system performing online software. The system implements the algorithm of MOORA and COPRAS techniques with an example of robot selection to test the applicability and validate multi criteria decision problem results. Results shows COPRAS and MOORA ranking of robots are very close to each other especially the 1st and second top alternatives. The software can be extended to the other decision making problems as well.

International audience; Applying agent-based modeling to simulate the evacuation in case of emergency situations is recognized by many research works as an efficient tool for understanding the behavior and decision making of occupants in these situations.In this paper, we present our work aiming to modeling the influence of the emotion and social relationship of occupants on their behaviors and decision making in emergency as in case of fire disaster. Firstly, we proposed a formalization of occupants' behavior at group level in emergency situations based on the social theory. This formalization details possible behaviors and actions of people in emergency evacuations, taking into account occupant's social relationship. The formalization will facilitate the construction of simulation for emergency evacuation. Secondly, we modeled the influence of emotion and group behavior on the decision making of occupants in crisis situations. Thirdly, we developed an agent-based simulation that took into account the effect of group and emotion on the decision making of occupants in emergency situations. We conducted a set of experiments allowing to observe and analyze the behavior of people in emergency evacuation.

Characterising tractable fragments of the constraint satisfaction problem (CSP) is an important challenge in theoretical computer science and artificial intelligence. Forbidding patterns (generic sub-instances) provides a means of defining CSP fragments which are neither exclusively language-based nor exclusively structure-based. It is known that the class of binary CSP instances in which the broken-triangle pattern (BTP) does not occur, a class which includes all tree-structured instances, are decided by arc consistency (AC), a ubiquitous reduction operation in constraint solvers. We provide a characterisation of simple partially-ordered forbidden patterns which have this AC-solvability property. It turns out that BTP is just one of five such AC-solvable patterns. The four other patterns allow us to exhibit new tractable classes., Logical Methods in Computer Science ; Volume 13, Issue 4 ; 1860-5974