Meet the AI Experts driving Innovation in TV Metadata

Nestled next to a little chateau and situated in a verdant park in the heart of Geneva is the headquarters of the European Broadcasting Union. Established in 1950, the EBU is the world’s leading alliance of public service media organizations. It has long been a driving force in digital transformation, as well as many societal and cultural initiatives from Europe’s public service broadcasters. The scope of its work is truly vast. The EBU operates a large number of expert groups that have impacted many aspects of the TV industry, not to mention having created the Eurovision Song Contest (nowadays with next generation audio, of course)!

As part of its technology and innovation program addressing media information management (MIM-AI), one of the key areas of activity today is the metadata and artificial intelligence work stream. Exploring a wide range of AI-driven technologies, this stream focuses on helping members enrich and enhance their technical and descriptive metadata to enable better experiences for their audiences. This also covers speaker identification, face recognition, fake news detection, sport highlights, etc.

It does this in a cooperative community atmosphere that brings together and showcases the international experience and best practice of academics, researchers and industry participants. The AI and metadata work stream has several components including Automatic Metadata Extraction (AME), Media Cloud Microservices Architecture (MCMA), Metadata Models and the Machine Learning (ML) Data Pool, as well as the Metadata Developer Network (EBU MDN), which is part of the EBU’s Strategic Programme on Production. This group has been growing steadily year on year both in terms of scope and number of participants.

The EBU Metadata Developer Network

The first Metadata Developer Network (MDN) workshop was organized in 2011. Since then, developers have been invited to come every year to the EBU to report on their most recent projects. Presenters were originally given an hour to go into technical details, but developers have been so keen to network at MDN that the allotted time had to be reduced to 30 minutes in 2019. MDN additionally hosts hands-on demonstrations where experts can try and challenge tools being presented by those who made them. From 2023 onwards, this event will be known as the Data Tech Seminar.

A review of the programs and papers shows how MDN accurately reflects the evolution of media information management within the broadcast community over the nearly last 10 years. Updates are made annually on all the fundamental aspects of metadata like modelling of data from commissioning to distribution.

Members and experts (especially metadata architects and implementers) are encouraged to come together to discuss key topics in metadata, participate in hands-on workshops and share best practice. Here you´ll gain access to tools, hints and examples to help you implement well-known metadata formats and standards.

Participants at the EBU Metadata Workshop in Geneva, June 2019. Foreground, Left to Right: Masanori Sano (NHK), Tormod Værvågen (NRK), Jean-Pierre Evain (EBU).

Every year, more and more participants are presenting new aspects of metadata and the marketing innovations to come in this field. Several European projects have participated to relay new information on their innovations and new fields of investigation. The use of semantic technologies and linked open data have been promoted, too.

The MDN is where the Audio Definition Model (ADM) was first publicly presented several years ago. The Interoperable Media Format (IMF) was also identified as one of the important changes to come. But the most striking example definitely is Artificial Intelligence (AI), which, although not named as such is a much over-used and misunderstood term. In fact, AI-driven and machine learning (ML) technologies have been present since the early days of the MDN (eg, automatic metadata extraction, natural language processing, audio and video analysis, semantic annotation, speech to text, automatic subtitling and captioning, automatic translation, personalization and recommendation, etc), before all the hype about cloud platforms took over.

The MDN is also where metadata workflows and systems can be discussed (metadata machine, smart data platform, advanced semantic MAM, service based architectures, metadata models, automation, optimization, etc.).

According to Jean-Pierre Evain, Principal Project Manager at the EBU, “Metadata is now recognized as the lifeblood of digital broadcasting. MDN has a bright future bringing experts together from larger horizons. No doubt that AI applied to news and sport will be at the core of the next edition among any other fascinating related research, development and production topics.”

The MDN offers program of in-depth presentations and demonstrations about current research and technical developments around metadata and artificial intelligence in broadcast and media. It’s the place to expose your opinions, problems and challenges as well as engage in lively and friendly discussions with your colleagues from all around the world. Topics include, but are not limited to:

o Any topic related to metadata

o Artificial Intelligence

o (Micro) service-based architectures

o Agile workflows in production, archives and distribution

o EBUCore implementations

o EBU CCDM implementations

o AI and automatic metadata extraction (AME)

o Orchestration of microservices such as accessing AI tools across cloud platforms

o Multilingual challenges

o Vocabularies and taxonomies

o Data models and ontologies

o Content tagging

o User interfaces and metadata visualization

o Prototypes, case studies,

o Success stories and learnings (such as epic failures)!

Learn more about the Metadata and Artificial Intelligence Technology and Innovation work streams here, and find out how to contribute to the next Data Tech Seminar here.

Updated: December 2022