Overall Objectives
Scientific Foundations
Application Domains
New Results
Contracts and Grants with Industry
Other Grants and Activities

Section: Other Grants and Activities

National Initiatives

ANR project Semim@ges

Participants : Zein Al-Abidin Ibrahim, Patrick Gros, Emmanuelle Martienne, Sébastien Campion.

Duration : 27 months, starting in January 2007. Partners: Orange Labs, TDF, Kersonic, Telisma, CAIRN team.

The project is devoted to TV data exploitation and repurposing. Two main applications were considered: TV news analysis, and TV streams structuring. TexMex project-team was mainly involved in the second one. The aim of our work was to structure automatically long TV streams in more usable units like programs or non-program sequences, exactly like it was done in Xavier Naturel's thesis, but with no a priori manual annotation. A method was developed based of the detection of repeated segments and on an automatic unsupervised classification of the repeated sequences. The Semim@ges project was demonstrated during the NEM Summit in Saint Malo, France, in September 2009.

ANR project ICOS-HD

Participants : Ewa Kijak, Joaquin Zepeda.

Duration: 4 years, starting in January 2007. Partners: University of Bordeaux 1, CNRS-I3S.

This project concerns scalable indexing and compression for high definition video content management. Recent solutions for achieving high-quality compression of images/video resulting in scalable bit streams. The objective of the project is to propose new solutions of scalable description to facilitate editing, manipulation and access of HD contents via heterogeneous infrastructures. TexMex project-team is involved on the study of new signal representation amenable to both compression and image description, as well as descriptor adaptation for image retrieval in large databases.

ARC INRIA RAPSODIS: Syntactic and Semantic Information-Based Automated Speech Recognition

Participants : Camille Guinaudeau, Gwénolé Lecorvé, Pascale Sébillot.

Duration: 2 years, starting in February 2008. Partners: Metiss , Parole , Talaris project-teams, CEA-LIST/LIC2M.

This project aims at improving automatic speech recognition (ASR) by integrating linguistic information. Based on former work by S. Huet concerning the incorporation of morpho-syntactic knowledge in a post-processing stage of the transcription, we experiment, together with our partners, the deep insertion of automatically obtained semantic relations (especially paradigmatic ones) and syntactic knowledge within an ASR system.

In 2009, the objectives of the project were extended to include semantic knowledge acquisition and the use of such knowledge for spoken document processing in addition to speech transcription. In this extended framework, we have worked on corpus-based acquisition of semantic relations for topic segmentation of spoken documents. We compared various classical methods for relation acquisition and measured their impact on out topic segmentation system.


Logo Inria