Search

International audience; This paper investigates the development of a formal model in order to analyse online hate in French corpora. Relevant concepts are identified by exploiting several sources: the cognitive foundations of the appraisal theory, according to which people's emotional response are based on their own evaluative judgments or appraisals of situations, events or objects; a linguistic model of how different kinds of modalities applied to predicative contents are expressed in textual data; several definitions of a hate speech. Based on those inputs, a formal model was developed to describe online hate speech. The model highlights different categories of hate targets and actions, and emphasizes the importance of context for online hate detection.

Exploring unexploited but newly digitized resources to find relevant information is a complicated task due to the amount of available resources. Thanks to the ANR project CIRESFI, the most important resource for the Italian Comedy of the 18th century, is a set of accounting registers consisting of 28,000 pages. Information retrieval is a long and complex process that requires expertise at every step: detection and segmentation in paragraphs, lines or words, features extraction, handwriting recognition. Systems based on deep neural networks dominate these approaches. The major issue is the need of a large amount of data to achieve their learning. However, the registers of the Italian Comedy have no ground truth. To overcome this lack of data, we explore approaches that involving transfer learning. That means using heterogeneous labeled and available data, with at least one common feature with our data to drive the systems, and then applying them to our data. All of our experiments have shown us the difficulty of carrying out this task, each choice at each stage having a strong impact on the rest of the system. We converge on a solution separating the optical model from the language model in order to achieve independent learning with different available resources and joining together thanks to a projection of the information into a non-latent common space.; La tâche d'exploration dans des ressources inexploitées mais nouvellement numérisées, afin d'y trouver des informations pertinentes, est complexifiée par la quantité de ressources disponibles. Grâce au projet ANR CIRESFI, la ressource la plus importante, pour la Comédie-Italienne du XVIIIe siècle, est un ensemble de registres comptables constituée de 28 000 pages. L'extraction d'informations est un processus long et complexe qui demande une expertise à chaque étape : détection et segmentation, extraction de caractéristiques, reconnaissance d’écriture manuscrite. Les systèmes à base de réseaux de neurones profonds dominent dans l'ensemble ces approches. Le problème majeur est qu'ils nécessitent d'avoir une grande quantité de données pour réaliser leur apprentissage. Cependant, les registres de la Comédie-Italienne ne possèdent pas de vérité terrain. Pour palier ce manque de données, nous explorons des approches pouvant opérer un apprentissage par transfert de connaissance. Cela signifie utiliser un ensemble de données déjà étiquetées et disponibles, possédant un minimum de points communs avec nos données pour entraîner les systèmes, pour ensuite les appliquer sur nos données. L'ensemble de nos expérimentations nous ont montré la difficulté de réaliser cette tâche, chaque choix à chaque étape ayant un impact fort sur la suite du système. Nous convergeons vers une solution séparant le modèle optique du modèle de langage afin de réaliser un apprentissage indépendant avec différents types de ressources disponibles et se rejoignant grâce à une projection de l'ensemble des informations dans un espace commun non-latent.

Exploring unexploited but newly digitized resources to find relevant information is a complicated task due to the amount of available resources. Thanks to the ANR project CIRESFI, the most important resource for the Italian Comedy of the 18th century, is a set of accounting registers consisting of 28,000 pages. Information retrieval is a long and complex process that requires expertise at every step: detection and segmentation in paragraphs, lines or words, features extraction, handwriting recognition. Systems based on deep neural networks dominate these approaches. The major issue is the need of a large amount of data to achieve their learning. However, the registers of the Italian Comedy have no ground truth. To overcome this lack of data, we explore approaches that involving transfer learning. That means using heterogeneous labeled and available data, with at least one common feature with our data to drive the systems, and then applying them to our data. All of our experiments have shown us the difficulty of carrying out this task, each choice at each stage having a strong impact on the rest of the system. We converge on a solution separating the optical model from the language model in order to achieve independent learning with different available resources and joining together thanks to a projection of the information into a non-latent common space.; La tâche d'exploration dans des ressources inexploitées mais nouvellement numérisées, afin d'y trouver des informations pertinentes, est complexifiée par la quantité de ressources disponibles. Grâce au projet ANR CIRESFI, la ressource la plus importante, pour la Comédie-Italienne du XVIIIe siècle, est un ensemble de registres comptables constituée de 28 000 pages. L'extraction d'informations est un processus long et complexe qui demande une expertise à chaque étape : détection et segmentation, extraction de caractéristiques, reconnaissance d’écriture manuscrite. Les systèmes à base de réseaux de neurones profonds dominent dans l'ensemble ces approches. Le problème majeur est qu'ils nécessitent d'avoir une grande quantité de données pour réaliser leur apprentissage. Cependant, les registres de la Comédie-Italienne ne possèdent pas de vérité terrain. Pour palier ce manque de données, nous explorons des approches pouvant opérer un apprentissage par transfert de connaissance. Cela signifie utiliser un ensemble de données déjà étiquetées et disponibles, possédant un minimum de points communs avec nos données pour entraîner les systèmes, pour ensuite les appliquer sur nos données. L'ensemble de nos expérimentations nous ont montré la difficulté de réaliser cette tâche, chaque choix à chaque étape ayant un impact fort sur la suite du système. Nous convergeons vers une solution séparant le modèle optique du modèle de langage afin de réaliser un apprentissage indépendant avec différents types de ressources disponibles et se rejoignant grâce à une projection de l'ensemble des informations dans un espace commun non-latent.

International audience; Among the computational intelligence techniques employed to solve classification problems, the fuzzy rule-based classification system (FRBCS) is a popular tool capable of building a linguistic model interpretable to users. However, it may face lack of accuracy in some complex applications, by the fact that the inflexibility of the concept of the linguistic variable imposes hard restrictions on the fuzzy rule structure. In this paper, we extend the fuzzy rule in FRBCS with a belief rule structure and develop a belief rule-based classification system (BRBCS) to address imprecise or incomplete information in complex classification problems. The two components of the proposed BRBCS, i.e., the belief rule base (BRB) and the belief reasoning method (BRM), are designed specifically by taking into account the pattern noise that existes in many real-world data sets. Four experiments based on benchmark data sets are carried out to evaluate the classification accuracy, robustness, interpretability and time complexity of the proposed method.

International audience; The oral entry of texts (dictation machine) is an important field of application for automatic speech recognition, and the first existing prototypes are encouraging the development of such applications. However, some research effort has to be carried out. The facilities that have to be improved, include the size of the available vocabulary and the utterance style (continuous or isolated speech). We are working on such improvements. In this paper, we propose a general overview of the MAUD project currently under devlopment. It is based on four major interacting components: an acoustic phonetic-decoder, a lexical component, a linguistic model and a user interface. MAUD uses a vocabulary made up of the most commonly used words in French and contains about 37000 entries extracted from BDLEX. We discuss first the main features of a dictation machine in the general framework of speech recognition. We demonstrate that the lexical component is one of the most important of our system, intervening as it does at a number of different levels: phonetical, phonological, morphological, syntactic, semantic, etc. We also describe our linguistic level which is probabilistic, based on bi- and tri-class frequencies as well as using a set of grammatical rules

International audience; The oral entry of texts (dictation machine) remains an important potential field of application for automatic speech recognition. The RFIA group of CRIN/INRIA has been investigating this research area for the french language during the past ten years. We propose in this paper a general presentation of the present state of our MAUD system which is based upon four major interacting components: an acoustic phonetic decoder, a lexical component, a linguistic model and a user interface.

International audience; It is proposed a medical knowledge representation of inflammatory protein variations involving complex relationships, frequently encountered in Internal Medicine. A linguistic model has been represented by a fuzzy set pattern expressing relative variations of serum proteins levels. Weights have been introduced into this pattern to translate relative importance among proteins. Three indexes or measures (possibility, necessity and truth-possibility) have been used for pattern matching purposes. Finally, a separating power yielding non fuzzy partitions has allowed to assign the corresponding diagnoses to patients (over 160 cases).