RTVE's knowledge graph

The fact that ontologies are a knowledge representation model independent of any system is what makes it possible to represent and consolidate data from diverse and heterogeneous systems in a knowledge graph. RTVE Grafo is an advanced data structure that organizes and relates information in a more intuitive and efficient way. This graph has been populated with the contents of RTVE Play, which allows a better organization and access to the vast amount of RTVE audiovisual material.

The RTVE Knowledge Graph, RTVE Grafo, is a representation system of the set of its contents and digital resources that understands facts related to programs, audiovisual contents, seasons, genres, themes, as well as any object potentially linked to them. When we say that it is a system that “understands” we must assume that it is a system written in a technical language that makes it possible for machines or systems to “understand” and correctly treat the set of entities to which we have referred in order to collaborate with people in their processes of interrogation, information retrieval and knowledge discovery.

Semantic annotation and population of the RTVE graph

Probably the most outstanding and far-reaching result of this digital project has been the consolidation of the contents from the Spanish Public Radio and Television into a large unified knowledge graph, extensible, expressive and interrogable by machines and people, making it easier for users to retrieve these resources according to any interest or intention.

For the consolidation of all RTVE Play data in the unified knowledge graph it has been necessary to design and develop a synchronization process that collects online the data from RTVE systems and annotates them semantically according to the defined EBUCorePlus-based ontology and adopted term vocabularies (as is the case of ESCORT 2007 - EBU System of Classification Of Radio and Television Programs), representing them in the form of triples (predicative sentences with the form subject+predicate+object) and depositing them in the semantic store (graph database) that is at the heart of RTVE's new semantic AI platform.

For the correct semantic annotation it has been necessary to perform a data alignment of the existing contents in the RTVE databases with the classes and attributes defined in the RTVE ontology and in some cases to improve the metadata of such contents in the source systems. A crucial objective of the RTVE Grafo project was to improve this metadata, that is, the way in which RTVE contents are labeled and described. A more accurate and detailed metadata facilitates the search and access to the information. From an internal point of view, the project aimed, therefore, at developing a fine-tuned system for the annotation and semantic representation of contents which would shorten the distance between RTVE and the varied set of audiences to which a public institution has to address and for which it has to speak. For this purpose, and beyond its public use. RTVE Grafo is used to annotate, organize and present the information in a meaningful way, gathering, for example, in the file of each content all the relevant information related to it.

The RTVE knowledge graph integrates some 2,000,000 digital resources, 26 million entities, some 85 million relationships between these different objects and entities, and 167 million triples, which are used to understand the meaning of the term that the user enters in the search; but also to offer a system for exploring the collection and, in general, all the resources, based on a faceted search engine, among other utilities, which allows the user to have all the possible ways of browsing this set of entities. The ontology and the RTVE Grafo will make it possible to represent the contents in a more precise, detailed, exhaustive and expressive way, and will facilitate more natural and conversational forms of relationship between users and the same.

In short, the exploitation of a knowledge graph of these characteristics will make it possible in the future to hybridize it with other artificial intelligence technologies in order to develop advanced services for different groups of users. This new way of being present on the Web aims, in short, to create and use the knowledge base of the Spanish audiovisual heritage in an intensive and efficient way.

Knowledge graphs represent the structure of reality and the mode of operation of our cognition.

RTVE Grafo wants to be useful to people and their demands for knowledge expressed through their questioning processes. To this end, the project has transformed the massive data and knowledge of RTVE Play into fast and accurate answers to complex questions in a scenario that assumes the need for the answers to be explainable, using artificial intelligence based on the emulation of human-like reasoning (semantic reasoning) operated with high-performance knowledge graph technology.

The RTVE Grafo knowledge graph is operated by AI for some specific purposes, such as the unification of the thesauri of the RTVE Archive, a key project for the enrichment of the metadata of its contents.