This research project approaches documentary photograph as a tool and a medium in order to analyze its circulation throughout three areas of uses, and to enhance its contribution in the making process of heritage. The issues raised by the researcher in charge o the project, Cecile Tardy, expert in information and communication sciences specialized in the area of the mediations of the inheritance, have the ambition to open a new field of research that combines communication through a scientific point of view, mediations of inheritance and photographs. The team brings together researchers able to cooperate around the set of themes of the photographs of the inheritance by bringing their field of interest together (devices of circulation and uses, logics documentary and semiotics of the image). The team is made up by a limited number of researchers, who have already worked on issues related to documentary photographs (as tourism, local history, literature, landscape, museum). We call "documentary photograph" a photograph produced within a scientific framework, but also socially requalified. We assume the stake of a documentary use of photograph is to get together two frames: 1. a thought and artistic process on which the content of the image is built and 2. a communicational point of view through which the evolution of that image can be seized as a media object. Thus, the photograph can be considered as a tool that allows a public made by experts to access scientific knowledge, or a media object open to a wider audience. And therefore, one can say that the photograph is put into circulation within different contexts of uses. It is by giving the photograph the double status of tool and medium that we consider its organizing function through a plurality of uses related to sciences, culture and institutions. This point of view on the photograph is relevant and meaningful when the heritage-building process is in stake, and moreover because this process requires at the same time a constructing of knowledge by the use of images and the establishment of a symbolic and social communication system (change of status of the objects, of the ways they are referred to and of the logics of acting). Thus, the photographs are chosen not because of what they represent, but because of their double relation to : an heritage-building process and a process of reappropriation in different areas of practices. During the study, our objects will consequently be determined by selecting devices that use photographs of the heritage and that are carrying logics of uses. Answering these problems requires renewing the most common standpoint towards a photograph in terms of piece, archive, speech or genre. The point of view on photograph as a tool in studying presents the major advantage : by reintroducing its materiality as a tool, it allows a study on photograph as a text opened to its context of practice and interpretation. In order to think the space of circulation of photograph, it must be considered, not only in its significance, but also in its social practices. Thus, thinking over photograph as a tool is defining it through three kinds of mediation : technical, textual and practical. These photographic mediations, involved in the heritage-building process, will be in the very heart of this study. In the 'heritage-building process', heritage object is a project, a part of heritage that is always to come. Photograph does no longer represent only heritage, but makes it occur by putting it in communication devices which lead heritage to interpretation, to handling, to qualifying. This analytical prospect shifts away the function and status of the image in the scientific work, since as a tool and a medium, photograph is no longer on the side of the researchers and of the construction of their knowledge only. It authorizes and conditions new relations between the subjects, the objects, and institutions of the heritage field. And then the device of this social deployment for documentary photograph must be questionned. Once photograph is in a space of communication that is wider than its context of production, it enters into constant reinterpretation. If the scientific use of documentary photograph seeks to maintain its initial meaning, the cultural and institutional uses will exploit the possible margin of interpretation, though without disqualifying the documentary dimension, but by widening the circle of the actors who guarantee the meaning of the image. This project will insist on the analysis of the communication device which organizes the setting of visibility of photograph.

One of the fundamental topics in current linguistic theory is the definition of the scope of variability in normal speech. The limits of what is to be considered "normal variation" can be determined by studying pathological variation. To do this, we need to know more about pathological speech. The aim of our project is to explore the large scope of speech variability through the description of acoustico-phonetics characteristics of dysarthric speech. Dysarthrias is a group of speech disorders resulting from neurological impairments of speech motor control. Substantial variations occur in dysarthric speech due to a deficit in the spatio-temporal execution of speech movements that affects different levels of speech production (respiratory, laryngeal and supralaryngeal). Our project aims at studying the scope of normal speech variation by characterising pathological varaition in dysarthric speech in phonetic-acoustic terms. The study will involve different types of dysarthrias through a large number of French-speaking dysarthric patients. A multidisciplinary team will bring together specialists from phonetics, phonology, clinical, speech and computer science engineering guarantee for the success of this project. The objectives are the following: (1) To identify and quantify the phonetic-acoustic characteristics of dysarthrias through an original procedure combining the potential of both manual phonetic acoustic analysis and methods of automatic speech processing (2) To assess the relevance of the criteria thus established by considering how well they are suited for distinguishing normal from dysarthric speech, for categorizing different types of dysarthrias, and for describing the longitudinal evolution of dysarthric decline; (3) To assess the robustness of the criteria considering the known constraints in normal speech in different speaking situations such as reading a sentence, telling a story based on picture interpretation, spontaneous speech. This project is somewhat held up by the fact that pathological speech recordings are widely scattered and difficult to come by. A preliminary stage of the project will therefore consist in putting in place a structured computer-based database of pathological speech. We currently have at our disposal 1,000 recordings of dysarthric patients put together by C. Chevrie-Muller between 1965 and 1997. This will be tied in with a major effort to harmonise terminologies between the different actors in clinical phonetics.

Looking for named entities on the Internet is becoming more difficult, especially since the rapid increase of multimedia data online. Even if the actual methods for named entities detection in textual data are quite mature, the detection on diverse multimedia objects seems to be much more difficult to model. Because of this difficulty, the data needed to find robust identifiers is much higher. The high cost of multimedia data annotation limits indeed the usage of statistical methods which have proved to be effective. On the other hand, the concepts (or entities) that could be looked for appear in different ways depending on the support (audio, text or video support). Thus, the conception of generic methods constitutes a major scientific challenge in the field of multimedia detection. The SuMAC project suggests exploring original learning methods to detect multimedia entities by using specific detection patterns. The usage of those patterns offers a unified framework to express different rules of combination. In this context, we will suggest low supervised methods to estimate the entity's signature for each media. Furthermore, we will develop active learning and cross-media co-learning methods aiming to diminish considerably the effort of supervised learning. All these methods will be evaluated in the frame of the Wikio web-portal that offers the possibility of initial structuring of data and running an evaluation under real conditions. The SuMAC project will cover these topics in the frame of a Fondamental Research Project with a duration of 36 months led by Laboratoire Informatique d'Avignon (LIA - Université d'Avignon), Eurecom Laboratoire and the companies Syllabs and Wikio.