Section: Other Grants and Activities
ACI masses de données Remix: Indexing Large Amounts of Data with Reconfigurable Chips
Participant : Laurent Amsaleg.
Duration: 3 years, starting in September 2003. Joint work with Symbiose and R2D2.
This project aims at developing new technologies to access very large databases like DNA or image databases. The core technology used in the project is based on FPGA chips.
ACI masses de données MDP2P: Managing Large Amounts of Data in P2P Systems
Participant : Laurent Amsaleg.
Duration: 3 years, starting in September 2003. Joint work with PARIS and ATLAS.
This project aims at studying the problem arising when managing lots of data on P2P systems like PC clusters. These problems are studied from a system point of view (memory management) and from an algorithmic point of view (parallelization of algorithms).
ACI masses de données DEMI-TON: Multimodal Description for Automatic Structuring of TV Streams
Duration: 3 years, starting in April 2005. Partners: INA, METISS project-team.
This project concerns the development of new techniques to index large collections of TV programs. INA records and index more than 50 channels 24 hours a day. As the number of available documentalists did not increase as fast as the number of channels to index, they have to rely on more automatic processes. The first need is to verify that the programs in the stream correspond effectively to what was announced in the TV program guide and to synchronize the stream with this program guide. In a second stage, some programs like news reports have to be indexed to the topics that were tackled by the program and which, of course, could not be announced in the program guide.
The first year of this project was dedicated to define places of collaborations between the three involved groups and to start the work in these three directions:
A first common work has been carried out to coordinate the effort done by the three partners to develop experimental platforms. This work will continue what has been achieved within the FERIA project, with new developments to take the hardware available at each partner's site;
Macrosegmentation of TV streams has been studied: the learning based approach of INA will be coupled with the signal based approach of TexMex ;
The interaction between speech transcription and natural language processing tools has been studied in a collaboration between TexMex and METISS.
At IRISA, a first platform to record and analyze TV streams has been developed (see Section 5.2 ). It is based on a Windows server in charge of recording and distributing the content and two Linux PCs in charge of processing tasks. It is now available to all the members of METISS, TexMex and VISTA through a simple web site.
Participation to National Working Groups
P. Sébillot is a member of the thematic network "Information and knowledge: discovering and abstracting" of the STIC department of CNRS.
P. Sébillot is member of the AS "Semantic Web" of the STIC department of CNRS.
P. Sébillot participates to AS "Text Mining" of the STIC department of CNRS.
P. Sébillot is a member of AFIA Café (Collège apprentissage, fouille et extraction).
P. Sébillot is a member of the working group A3CTE: Application, Learning and Knowledge Acquisition from Electronic Texts of GDR I3.
P. Sébillot is a member of the working group PRC I3-AFIA TIA (terminologie et intelligence artificielle)