Documents are not random sequences of sentences, but rather spans of texts are linked together into a discourse structure to make coherent and meaningful documents. Developing systems able to interpret documents and to extract structured information is a challenge in natural language processing that requires high-performing discourse parsers. However, performance are still low and current work does not make it clear where the problem lies. Moreover, studies focus on English monologues from news, which limits their use and their robustness. Finally, empirical studies rarely try to inform theoretical models for discourse. We propose a general framework here called multi-objective learning to tackle these issues. We will build systems seeking to achieve several goals as a way to provide robust systems for multiple languages, domains and modalities, to investigate data representation and problem modelling, to improve evaluation, and to shed some light on the theoretical divergences.

Mechanism Design aims at defining algorithms and procedures for organising the selection of some outcome with respect to the preferences of human and artificial agents. Selecting a good outcome is at first importance as these algorithms extensively regulate society: e.g., election, household allocation, refugee relocation or student selection; the key problem is that stakeholders may use strategy for taking advantage of the mechanism for their own benefit. The aim of this project is build up a consortium for an Horizon Europe submission. The intended call is the HORIZON-CL4-2024-HUMAN-01-07:combining the best of machine and human The submission topic will be be on how to use and extend formal verification techniques used for guaranteeing the quality and robustness of computer-based systems for designing good mechanisms regardless of the strategic aspect. Main benefit is that Mechanism Design will be redefined in terms of synthesis: human designers focus on the expected properties of the mechanism and, accordingly, the machine generates mechanisms well defined and explainable by design. Second benefit is, by challenging the formal verification techniques with societal problems, we will address scalability, user interface issues, make them progress and usable by a large audience.

The Structures project proposes to study new approaches for the creation of stylized animated films in connection with traditional artistic practices. Our environment is surrounded by images and films. Stylized images carry the visual identity and atmosphere of an artistic movement, an artist or a business. This is why artists need both great freedom and precise control over the appearance of the image to express their style when creating images or animations. Unfortunately, with the current tools and stylization techniques, there is still a lot of manual work to create animations that express the artistic intent of the content creator. This not only limits the quality and variety of styles, but also dramatically increases production costs. As part of our project, the term style design the visual appearance of the images constituting the animation. Our project answers the following questions: - How could the industrial production pipeline be better adapted to the artist to produce productions of a quality similar to that of handmade productions? - What are the motion information to describe and how to model this motion to guide stylization? - What is the motion to describe and how to model this motion to guide stylization? To answer these questions, the Structures project addresses the following challenges: - understand the stylization process followed by artists to create animations and study motion stylization models simulating this process; - define how to represent the motion of the input scene so that the stylization process has access to relevant information; - establish a vocabulary to describe existing styles; - define a style control mechanism adapted to artists. These challenges fall under two scientific disciplines: computer graphics and visual arts sciences. We propose a new approach associating researchers and practitioners in computer graphics and art sciences. The Structures project team covers the different skills and knowledge needed to conduct this research. The Structures project focuses on the artistic creation of hand-drawn animation with traditional tools such as pencils, pastels, painting. Even though every production of this type has its own production pipeline, we are convinced that common practices can be identified. This will allow us to understand how the motion information of an input scene is related to the style. The objective is to propose a model of representation of the motion of the geometry and the motion of the appearance. This model will guide the process of stylization. To this end, we need to formalize the drawing process and define stylization algorithms that implement this process. We will define an intermediate representation between the input scene and the stylization to guide the control of the style. This representation will be motivated by artistic practices and will provide new controls in accordance with the artists' vocabulary. The results will be a dictionary of styles covering artists' vocabulary as well as a stylization pipeline. We will develop stylization algorithms with improved control through the use of our motion representation structure. These results will have a direct impact on visual communication, for the creation of animation. We also hope to better understand the artistic process of creating stylized animations and have an impact on visual arts science.

Interactive training systems can embody so-called intelligent technologies: robotics, activity recognition, intelligent tutoring systems and artificial intelligence. For stroke rehabilitation, they cover specific objectives of a rehabilitation program. In order not to limit the intelligence of these systems to specific objectives, the TEAM UP project will provide a tool-based approach to the design and development of intelligent interactive training systems that know and recognize all the user tasks in an individual rehabilitation program and are able to adapt their behavior to the user's progress. Thanks to this project, Célia Martinie will develop her own field of research in the engineering of intelligent interactive training systems. She will rely on on her expertise in interactive systems engineering and on the health ecosystem in the Toulouse region.

Updating and maintaining hypervisors are tricky operations, but very common in data centers. Virtual machine (VM) migration is the main tool in the update process. VMs on servers are migrated to third-party servers, the hypervisor update is performed, and the VMs are retransferred. Although it is functional, this approach has many limitations: time-consuming because of migration, saturates network resources in the data center, and often requires a shutdown of client VMs. This project presents WalkIn, which is a platform for performing updates on main open-source hypervisors without the need for migration. WalkIn saves the running states of the VMs in memory, performs the update on the hypervisor, and restores VMs on the server all with a low VM interruption.