Your browser is out-of-date!

Update your browser to view this website correctly. Update my browser now


VRT: Speech recognition for automated annotation

The project AARON deals with automatic annotation tools for media archives through speech recognition, a face detector, keyword spotting and speaker clustering. Begun in 2013 by Belgium’s VRT research institute, it resulted in a prototype that offers a more effective way to perform search enquiries in archives without the usually labour intensive manual annotation.

VRT is exploring ways to make use of these extraction tools from the start of the production chain and to offer a personalised experience to end users. It also aims to make a smooth navigation of a collection of media material possible by tying up content visually and semantically.

“Based on the prototype we concluded that speech technology can be very effective in making archives accessible in a fast and adequate manner,” said VRT’s Evelien Van Malderen. “At our booth we are demonstrating the latest version of the prototype for radio news archive content.”

It is also exploring new ways to translate traditional documentaries online. At IBC it aims to simulate a documentary experience by setting up online platforms “in a tactile manner”.