RTL at the Spark+AI Summit

How Data is Transforming the Dutch Media Industry

Maurits van der Goes
Nov 7 · 4 min read

Amsterdam hosted “2.300 data scientists, data engineers, and global business leaders from 63 different countries” for this year’s Spark+AI Summit Europe. A three-day event with great keynotes, selected community talks, hands-on tutorials, and training sessions. The stage for the latest developments on powerful data tools like Spark, MLflow and Koalas. Sharing knowledge and meeting new people around the corner in our capital.

RTL Netherlands was present with multiple colleagues and data scientist Maurits van der Goes shared our work in the talk ‘How data is transforming the Dutch Media Industry’, part of the AI Use Cases track. It was a behind the scene look with five use cases, including technical details, drawbacks, and challenges.

30 years ago RTL started the first commercial broadcaster in the Netherlands. Pioneering in a traditional industry. Growing from one to multiple television channels. Launching international breakout hits Big Brother and The Voice. And, adding the biggest weather app and more online platforms. Today we reach 8,9 million people daily via television, have 779 million monthly views online and 2,3 million daily unique visitors on our digital platforms. This makes RTL a true technology and media company.

Consumers in control

Technology puts our consumers in control. They have more content available than ever, multiple devices to watch it, and they determine when to watch it. For RTL to help them discover, three questions are important for all our content:

  • content do we have?
  • is our audience?
  • are they watching?

Data is the answer to all of these questions. It allows us to deeply understand what our consumers love and go beyond our gut feeling. This is essential to create and improve our content in this very competitive market. Our challenge is to bring art and science together. To enable this the Data Intelligence department, with 22 colleagues from 12 different countries, is working on a wide range of data products throughout the whole company. And as part of RTL Group & Bertelsmann, we’re collaborating internationally. We build with e.g. AWS, Spark, Python, R, Jupyter, TensorFlow, Airflow, SQL and Kubernetes. Our approach is to use a mix of ordinary machine learning and deep learning.

5 use cases

With the pioneering energy of our colleagues, we are again redefining the Dutch Media industry. The five use cases that we shared in this talk:

  • On our news and video-on-demand platforms we’re using a variation of content-based recommendations, collaborative filtering, and deep learning approaches. For some use cases, simple Spark MLlib works fine, but for others, such as recommendations based on a sequence of items consumed (last watched) and when integrating visual features, deep learning methods are essential. We’re constantly testing our models’ performance using Bayesian A/B testing.
  • We believe that media is emotion and our mission is to tell stories that touch the mind & heart. Accordingly, detecting emotions and sentiment is very interesting to us. We use a multimodal approach to analyze emotions in video, text, and audio.
  • We combine scenarios and subtitles to identify items in talk shows. With Spark MLlib and deep learning methods, we can classify these items. This enables episode reporting and trend analysis and allows us to answer questions such as what subjects does our audience like?
  • Our content and advertising departments are highly interested in more accurate forecasting of television ratings. This used to be a manual process. We model based on multiple components, such as a long-term trend, yearly seasonality, weekly pattern weather information, and (sports) events.
  • We produce many trailers for our television content. They are ideal for building awareness and branding. However, this is a costly process. AI can support the analyzing, selecting, and combining of shots and scenes. Can you see the difference? Which trailer is created by a human or generated automatically? The audience had a hard time guessing it right.

Do you want to know the details of these use cases and RTL’s Data Intelligence team? Check out the YouTube video below.

Obviously, it does not stop here. Our goal is to deeply understand our consumer: the fan-centric approach, to tailor our television programs, online video and other content to the interests of the diverse Dutch audience. With the progress we are making, we are confident that the next breakout hit will be data-driven!

You can join us on this exciting journey! New ideas? Ready to challenge us? We’re hiring via werkenbijrtl.nl.

How Data Is Transforming the Dutch Media Industry (YouTube)
How Data is Transforming the Dutch Media Industry (SlideShare)

RTL Tech

RTL Technology Blog brings you stories from inside the RTL Technology Organization.

Maurits van der Goes

Written by

Data Engineer / Scientist @RTLNL & Student @TUDelft

RTL Tech

RTL Tech

RTL Technology Blog brings you stories from inside the RTL Technology Organization.

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade