Application to information searches
The information retrieval solution that we propose is a semantic search that allows the sought objects to be structured, this structure to be projected onto the knowledge base and sub-graphs that match it to be produced, as shown in Figure 4.13. This structure is established with the support of the graph editor (section 4.3) in the form of an RDF graph.
Figure 4.13. Representation of semantic queries
Semantic searches are always performed according to domain ontology. It allows the exploitation of RDFS/OWL2 reasoning capabilities to find objects whose descriptions do not exactly follow the precise outline of the query. Let us take the example of the following query, “Where is the Eiffel Tower?” knowing that geographic locations are connected among themselves by a single relationship: “is in”. In this way, a classic information search results in the single individual that is directly tied to “the Eiffel Tower”, namely, “the 7th arrondissement”. A semantic search engine, on the other hand, exploits the transitivity of this relationship and deduces the entire corresponding geographic hierarchy, namely, “the 7th arrondissement”, “Paris”, “Ile-de-France”, “France” and “Europe”.
In the framework of audiovisual content analysis and publication, semantic searches allow us to find:
- - media, media extracts and segments to which a precise description graph has been attached;
- - structuring objects like the analyses and description strata that respect a given annotation outline;
- - individuals and named entities whose description responds to different constraints expressed by the semantic query graph.