Single Sign-On and Recommendation System developed by Broadcasters for Broadcasters
The right content at the right time for the right person on the right device – as broadcasters are faced with competition from multinational online-giants and looking for ways to efficiently increase reach and relevance, the European Broadcasting Union is working with its Members to give editorial teams a set of powerful tools to deliver relevant content to audiences.
Keeping control over the algorithms driving content delivery is key for Public Service Media, not only to maintain editorial independence and impartiality, but also to live up to the trust placed in them by the public in today's data-driven world.
To that end, Public Broadcasters RTS, BR and others have joined up around the EBU's PEACH project (Personalization for EACH), a working group applying InnerSource practices to co-develop the required solutions.
The group is in charge of developing two digital products:
The first is a Single Sign-On system which allows audiences to personalize their media experience across several devices. Limited input devices are covered using the ETSI 103 407 Technical Specification. It allows the audience to authenticate on Connected Radios, HbbTV and Voice User Interfaces such as Amazon Echo.
The second product is a Data Pipeline, which empowers data scientist to explore data sets, quickly prototype recommendation algorithms, deploy them and then measure the impact. The algorithms are designed to respect the Public Service Media remit to inform, educate and entertain, and to avoid "filter bubbles" and other typical pitfalls of recommendation systems.
Currently, the system can make recommendations for all types of genres (text, audio, video) using four main parameters. “Trending” delivers popular content; “Content-Based Filtering” proposes similar content based on metadata; “Collaborative Filtering” attempts to deliver relevant content based on similar usage patterns seen in all users. It can be declined to expose user to a broader catalogue of content: “Diversified Algorithm”.
PEACH products are flexible and allow any data scientist to build bespoke recommendation algorithms and measure their performance without the need for specific data engineering skills.
In order to help you start using it, here is a set of tutorials which will walk you through the different features of the products:
Data scientist platform
The following recommendation algorithms are available in the Data Pipeline:
- 1 - Generic Content-Based Filtering
- 2 - Collaborative Filtering
- 3 - Diversified Algorithm
- 4 - Trending