stvd-kg: A Knowledge Graph for French Electronical Program Guides

Electronic Program Guides provide detailed information on TV programs, useful for offering user services or quality datasets. Unfortunately, their exploitation is limited by the absence of standardized semantics, text descriptions that are difficult to analyze and information that is often incomplete. We introduces here the knowledge graph stvd-kg (for the KG of large-Scale TV Dataset), which enriches a year’sworth of French TV EPGs by taking advantage of Wikidata and IMDb. This large KG (5.7M entities, 27.6M triples) contains 2.9M of publication events (corresponding to 1.7M television contents, 70k televisioncollections, and 84k named entities). To achieve this, we use an ontology adapted to TV content and its broadcasting, based on the EBU Coreontology, schema.org and RDF-schema. We also propose original mechanisms for cross-enrichment of information coming from named entity recognition and entity linking systems. Experimental evaluation of theknowledge graph stvd-kgbased on a ground truth shows its very high accuracy. The resources to construct KG stvd-kg is made publicly available at our Gitlab. The stvd-kg in Turtle format is attached below.