Back Back

Big data, a partnership between France Télévisions and the Centralesupélec digital lab

Open Innovation
Published on December 03, 2019

France Télévisions - Paris Digital Lab partnership: a win-win agreement. The Paris Digital Lab, the digital innovation program of CentraleSupélec School, has made available to France Télévisions three of its best talents to create decision-support tools with the help of Big Data, as part of a partnership. Seven weeks to solve a problem, that's the time allowed to succeed. Challenge accepted and achieved!

The objective? Introduce AI to analyze user behaviors

France Télévisions wanted to use Artificial Intelligence to analyze user behavior on all France Télévisions platforms. Thanks to its ability to analyze massive data, AI is indeed able to identify user characteristics and bring them together in audience groups in order to offer them a personalized experience. 

The purpose of the application at the Paris Digital Lab is therefore the creation of decision support tools with the help of Big Data (trend analysis, visualizations, reconciliations between data) that should eventually allow :

• The improvement of users' knowledge

• Better conversion rates for all content offered

• The discovery of unknown behaviors

"Currently, our product relies heavily on human capabilities to create these user segments from a database management system (DMP). As a result, France Télévisions is aware that it is "missing" various elements that need to emerge on our sites and applications. This is what the collaboration with the Paris Digital Lab is all about: enabling us to create data science models that will search for elements unseen to the naked eye among the mass of data we possess by bringing up the relevant information using algorithms, and then it is up to us to exploit them. " explains Valentin, one of the people in charge of monitoring this project at France Télévisions.

To do this, and with the support of internal teams, CentraleSupélec's student-engineers have tackled the creation of these tools using Machine Learning (ML) and Automatic Natural Language Processing (TALN).

What methodology was used to succeed in this challenge?

To do all this: a short time of 7 weeks. "In agile mode, with one-week sprints that allowed the France Télévisions teams to have a regular and continuous follow-up with the students," explains David, technical project manager, "This gave us a good framework and usable results. Daily calls facilitated a continuous and regular follow-up with the students to answer any questions."

30 gigabytes of daily data were made available to them over a two-month period. The start was done on their machines, locally, but as the volume grew, France Télévisions set up a project on their Big Data environment (Google Cloud Platform) so that their algorithms could be used directly in the Cloud and the calculations completed in a few days.

These devices have made it possible to create market segments, assess their relevance at the time and allow them to evolve over time.

The France Télévisions teams are delighted with the experience.

Assessment of the partnership

To learn more about the possibilities of AI for the media, find our AI Usage Mapping.

 
Written by FranceTV Lab
Published on December 03, 2019

Join our
next
testing session!

Final test before launching automatic subtitling of franceinfo

Closed captions are produced everyday for the French public news channel, from 6 to 9 AM and from 9 PM to midnight. In addition, on France TV website and applications, franceinfo will receive automatic subtitling thanks to an Artificial Intelligence during the rest of the day.

We'll keep it, tweak it, or
remake it—you decide!

Other related articles with #AI

Ultime test avant le lancement du sous-titrage automatique de franceinfo

Final test before launching automatic subtitling of franceinfo

Closed captions are produced everyday for the French public news channel, from 6 to 9 AM and from 9 PM to midnight. In addition, on France TV website and applications, franceinfo will receive automatic subtitling thanks to an Artificial Intelligence during the rest of the day.

MédIAs, nouvelle génération : le cahier de tendances de Méta-Media pour comprendre l’impact de l’IA générative sur l’Information
News

Next-Generation Media: Meta-Media's trend book for understanding the impact of generative AI on information

Just over a year ago, global tech giants, led by Mark Zuckerberg before quietly backtracking in March of this year, promised us a new future that would allow everyone the leisure of venturing into alternate worlds, adopting desired identities, and connecting to the world like never before through the Metaverse. Since November last year, a completely different innovation has been shaking up the world of tech and beyond, as this technology is much more democratizable (and already democratized) than the Metaverse. As you may have easily guessed, we are referring to generative AI!
Authentifier ce que l’on voit pour mieux donner à voir : c’est l’ambition lancée pour un nouvel outil de fact-checking
News

Authenticating what we see to better give what to see: this is the ambition of a new fact-checking tool

Finding the truth in the false, or the false in the truth, is a situation we can all be confronted with. At a time where information is consumed on social networks and platforms, it is sometimes difficult to determine what is right. Fact-checking appears to be an activity that should be developed through a new shared tool.