The goal of the PERCOL proposal is to participate to the REPERE evaluation program by building a consortium including partners from the industry (Orange Labs) and academic labs (LIA, LIF, LIFL). The proposal is structured in 3 mains sections: . 1) Using existing technology to cover the different modalities and tasks included in the REPERE evaluation program This section is based on the use of modules already existing within the partners of the PERCOL project: - Speaker segmentation and recognition - Automatic Speech Recognition - Optical Character Recognition of text within video - Named Entity recognition - Visage recognition - Person recognition 2) Developing research, within the REPERE framework, allowing combining different scientific fields (audio, text and video) and moving forward the state of the art in each of them for performing the main task of REPERE. The two main scientific directions that are going to be addressed are: - Multimodal fusion algorithms for automatic person recognition in video broadcast. - Improving information extraction from speech and images thanks to a combine decoding using both modalities to reduce decoding ambiguities. 3) Introducing ergonomic analysis in the definition of the task. Thanks to tests with Orange Labs video retrieval platform users, we will work on the metrics that are going to be defined for the REPERE evaluation program.