Abstract

One of the first and most successful areas of applications of Data Science is today developing around Modern Media services.

In this presentation I will illustrate typical applications that the European Broadcasting Union (EBU – http://tech.ebu.ch ) has been and is developing together with its members and are today used by Public Media Broadcasters in Europe. Starting from an application already deployed and used operationally, to new initiatives and developments aimed at product and services to be operationally deployed in the near future.

The first area of application is centered around the use of an AI based Recommendation and Personalization set of tools allowing the personalized delivery of media content to individual users: The right content at the right time for the right person on the right device. In particular the structure of the PEACH product, Personalization for EACH https://peach.ebu.io/ will be presented.

The second area of application is centered around the use of an AI based set of tools for language management. This set of tools will provide Speech to Text conversion, Text to Text translation and Text to Speech synthesis. The vision is here to allow any European Citizen to access any content produced by any media provider in his mother language. In particular the structure of the Eurovox initiative will be presented and a set of practical applications illustrated.

The third area of application is focusing around the use of Data Science tools to support News production and exchange, identification of Fake News and delivery of News services with a guaranteed quality targeting any citizen population. The European Broadcasting is today operating the largest European News Exchange service and works together with its members to develop new tools to always guarantee the best quality of service and the highest reliability and news independence.

Finally, Automatic metadata generation will be presented. This involves deep learning and machine learning algorithms design. The metadata automatically generated are used to value the media archives content, but it also opens the way of real-time content enrichment and hyper-distribution (distribution of content on multiple media delivery systems). Typically, these automated metadata are used for sports events distribution on social networks and in particular on facial identification and automatic highlights classification in sport videos.