Jürgen Grupp (SWR)
“Moving from data to knowledge is the way forward in a digitized organization.” So says my colleague Alexandre Rouxel, senior project manager for Data and AI at the EBU. “Data are not the oil of the future; they do not disappear. They can be linked to generate knowledge in a cumulative way that can be further exploited by people and machines.”
Knowledge can be defined as the meaningful connections between dots of information. People collect knowledge to gain insights to help better solve problems. This is how our brains work. Similarly, machines represent information as data, and machine learning (ML) can link that data to generate knowledge.
In the world of semantics, machines can represent knowledge in a graph. The knowledge graph is a data structure describing entities and their relationships by means of nodes and arrows (see graphic). Every arrow, its source node and its target node form a ‘triple’. A triple represents a simple statement like: ‘Mr. Spock is a character in Star Trek’ or ‘William Shatner plays James T. Kirk’.
The meaning of the nodes and arrows as well as the rules for building the graph are often defined in an ontology. So, we can understand ontologies as a specific type of data model.
While ontology design is a field of AI, it is executed by experts from the business domain. Good ontologies are business-oriented and technology-independent, making them interoperable by nature.
New metadata standard
EBUCorePlus is an ontology for media enterprises. It is defined by EBU Members for the media community. It follows up on two long-standing EBU ontologies: EBUCore and CCDM (Class Conceptual Data Model). The two were merged and thoroughly revised. The result is EBUCorePlus, the new standard that can fully replace its predecessors. It inherits both the long-lasting reliability of EBUCore and the end-to-end coverage of the media value chain of CCDM. EBUCorePlus is specified using the ontology web language and therefore strictly semantic.
Moreover, EBUCorePlus serves as a plug and play framework. It can be used out of the box, either in its entirety or just a subset of its elements. But it may also be adapted and extended to enterprise-specific needs. Especially for system integration tasks and defining requirements, projects benefit from EBUCorePlus as a business – not technology – oriented language.
For example, vendors that deliver data to media enterprises approached us at the EBU booth during IBC2022. They are interested in using EBUCorePlus as the standard delivery format for their services. This format will come at no extra cost for media enterprises, whereas any further transformation towards a customer-specific data format would come at a cost.
Moving from classical databases to knowledge graphs requires extending development skills from entity-relationship models to ontologies, from tables to triples, and from SQL to SPARQL (a query language for graph databases). This is where the EBU’s CorePlus Demonstrator Kit (CDK) comes in handy.
The CDK can be compared to a flight simulator: a developer can explore all the knobs and switches, playing around and practising all desired manoeuvres, but in a safe environment. The CDK is available in three deployment modes: cloud, hybrid and on- prem. It contains a graph database, populated with the EBUCorePlus ontology and some sample data. The CDK is free to use after registration (via email@example.com), and the cloud version does not require any installation.
So, fasten your seat belt and take off for your next destination: EBUCorePlus!
This article first appeared in issue 54 of tech-i magazine.