Véronique Demilly (France Télévisions)
As a telecommunications engineer working in the audiovisual sector for over 20 years, mainly in the technical field, I have observed that we are not always able to make full use of our data. One set of data is usually collected or created and intended for one use only. If you need it for another use, you must put in place process to adapt it. There is no interoperability: one dataset, one use.
I have also seen cases where a set of data that would be extremely useful for a specific application was not correctly stored, once it had been used. It was therefore difficult to retrieve it and sometimes even necessary to recreate the data from scratch: one dataset, used once.
The broadcast sector and more generally the audiovisual sector have been faced with a new challenge in recent years: the emergence of a large number of new partners. These range from new platforms, on which you must make your brand visible, to new subcontractors to support the development of AI skills. For each new partner, it is necessary to undertake a technical project to connect systems and to begin to work together: one partner, one project.
This is no longer sustainable. It’s a matter of operational efficiency and even of financial efficiency.
That is where TEMS comes in.
Introducing data spaces
I discovered the concept of data spaces thanks to the Gaia-X initiative. A data space is a distributed system defined by a governance framework that enables secure and trustworthy data transactions between participants, while supporting trust and data sovereignty. A data space is implemented by one or more infrastructures and enables one or more use cases.
As per Francesco Bonfiglio, former CEO of the Gaia-X international association, in a data space, you connect once, and you can access large amounts of data from various stakeholders. This is exactly what we need in order to break down data silos.
One of the main objectives of a data space is to facilitate data holders keeping control over their data and making it easily usable by different systems.
As far as the content production ecosystem is concerned, the first pitfall to be eliminated is the loss of data – in our industry we usually refer to metadata – during the production process. These metadata shall then be inextricably linked to the audiovisual work and thus easy to retrieve. Not only shall they be linked but they shall also be qualified: metadata are stored only if the appropriate person created and verified them.
All data and metadata shall be reusable, exploitable and finally exploited, with the generated value falling mainly into the pocket of the data owner.
Of course, once you have qualified data associated with an audiovisual work, you can imagine and create new ways of exploiting it. It is possible to improve visibility and findability or retrievability, combat piracy, and many other use cases.
Finally, generative AI must not make money on the back of results from algorithms trained on our data, on our content, without sharing the money with European rights-holders and creators. I am increasingly convinced that a secure environment like that offered by a data space can help track unwanted access to data and can control the respect of opt-outs or the contracting of data usage for training algorithms in a win-win manner.
Data spaces are not only about technical interoperability but also deal with automated contracts, clearing-house services, automated compliance checks, new business models, trustworthy governance, and more. This is what you can expect from TEMS.
This article first appeared in issue 58 of EBU tech-i magazine. Véronique Demilly will speak about media data spaces at the EBU Production Technology Seminar 2024.