The world of media is strongly impacted by the digital revolution: means of production, publishing, and broadcasting broaden and users' practices evolve, changing on the way established economic rules. The CrossMedia Observatory aims at building processes, tools and methods to analyze and better understand the challenges and the evolution of the media sphere. Our objective is to create an integrated system allowing the analysis of a large range of information sources, web sites (blogs, Press, TV and Radio websites), newspapers, Radio and TV media streams in a large period of time (12 to 18 months). More precisely, we want to detect, identify, monitor, measure, analyze, and study the spread of media events and their derivatives. The CrossMedia Observatory project is driven by the needs of social sciences research and of information protagonists. This strong collaboration between scientific research and social science aims at developing new concepts and analytical tools able to characterize the evolution of the media sphere. The project gathers partners from several research domains: linguistic and text analysis, transcription, computer vision, search engines, and visualisation. Each scientific partner brings not only their expertise in their field, but also makes available to the project high performance software. Therefore the project may focus on the main technological bottlenecks and challenges: formalisation of the crossmedia objects with relations and event concepts; transcription enrichment from cross media resources; discovery and tracking of events and dynamic visualisation. Additionally those are processed on very large scale data. The strengths of this project lie in the synergy between IT and social science research, the development of high performances tools, together with resource diversity and volume. The ambition of the TransMedia Observatory is to make the relationship between the Internet, press, radio and television broadcasts more intelligible, viewable and searchable at a macro-level of analysis, The automatic tracking and monitoring of media entities and events at a variable temporal granularity will facilitate this. If the proliferation of media sources that broaden users' practices is a reality, does it ensure nevertheless the information plurality? Who produced the information? Who is the source? These questions are a matter of general interest for the society of today and tomorow.

Cartec is an industrial research project which focuses on the analysis of the temporal evolution of hypermedia information data sets that evolve in a random way, in order to transform them into a structured corpus. This transformation raises three main issues, forming the three main activities of the project: 1. The ability to crawl a data set, whose evolution is unknown and variable amongst its documents; 2. How to index information so as to be able to search and navigate within the corpus, as well as to analyze its evolution; 3. Information storage: in order to encode the temporal variations of a result set and to display them at query time. Taking into account the technical features and constraints of this issues, it's essential to make the different modules work in an asynchronous way, with monitoring and synchronizing features and a global configuration process. For example, the document analysis would lead to define the document crawling frequencies. Cartec will apply its result in the global frame of the temporal indexation of the World Wide Web. This context enables to focus on applications in the field of large scale information intelligence. Finally, Cartec collaboration process will privilege short cycles. This feature enables concrete advances - and injection of these advances in the main product during the main project life cycle. This operational mode takes profit from the chosen duration (36 months) to build a long term vision and to plan short terms results according to the partners ambitions and industrial projects.