The AIDAR project
The project intends to maximize the automation of the archiving process, through the use of speech recognition, automatic text segmentation and classification techniques. A broadcast news transcription system (Speeral) has been integrated and tailored to the users needs, using specific language models and additional with segmentation classification algorithms.
An interface is being developed in order to provide a tool for archivists, that will help them creating semi-automatically meta-data for Radio and TV archives.

Budgets & Duration
The AIDAR project has a total budget of 720.420 € over 24 months.

Participants, Roles in the project & Deliverables

ULB (Belgium), Machine Learning Group: The group brings its expertise in the project for the development and implementation of the segmentation and classification algorithms. www.ulb.be
TITAN (Belgium): Has defined the data architecture X-AXIS for the project. Responsible for dissemination. www.titan.be
Voice-Insight (Belgium): Project leader, integration of the broadcast news transcription system; development of specific language models; development of a user interface for archivists. www.voice-insight.com
LIA (France): The « Laboratoire d’Informatique de l’Université d’Avignon » has developed a broadcast news transcription system called “Speeral” used for this project. They also bring the expertise for developing additional functionalities within this platform. www.lia.univ-avignon.fr/
RTBF (Belgium): Help in designing the user needs for an archivists interface, provide the database (samples and related meta-data, thesaurus) for developing the prototype of the application. www.rtbf.be
SKEMA (France): Specification and development of X-AXIS. www.skema.fr

Status
The AIDAR project ended Q2 2007.

More information

Top