Suggestion of property values
The objective of a model graph is to guide the user in his annotation task. This includes precise suggestions of the most appropriate ontological elements. The OWL2 restrictions are transformed into Prolog and SPARQL queries in order to find the classes of the ontology, the individuals already created in local contexts and the shared entities in the triplestore.
Figure 4.12. Suggestions of named entities
In this way, for each property of an editable entity (note, named entity, etc.), the annotation system suggests the compatible classes and individuals. In Figure 4.12, a list of countries has been suggested by the tool, as the restriction set on the property “localization” of a “subject” stipulates that the values of this property must be countries of the world.
Opening on the web of data
Manual annotation is the more reliable method to guarantee high quality data and is often adapted by businesses/institutions that worry about this quality. However, the human annotator is naturally subject to error and slightly less trivial annotations generally require expertise and familiarity with very diverse domains, otherwise the annotator can spend a considerable amount of time searching for the relevant, reliable information he needs. Historically, the annotator has searched for this information in encyclopedic books and manually completed his analysis.
With the appearance of the Web, this practice has evolved towards online research where the annotator can consult encyclopedic websites like Wikipedia. In recent years, the Web has seen significant progress in the representation and accessibility of data where information searches are done more often on better-structured knowledge bases that offer new data processing capabilities. For example:
- - the interconnection and binding of these bases allows the problem of data incompleteness to be overcome;
- - this binding also allows content from different origins to be connected among themselves, thereby improving the visibility and accessibility of these contents;
- - the structuring and formalization in the form of semantic graphs promotes reasoning of the data.
The idea is thus to take advantage of these abilities to find and collect the data the annotator needs online; this allows him to accomplish his documentation task quickly and efficiently. The semantic annotation system that we propose here exploits the same mechanisms provided by the graph editor (presented in section 4.3.6) to access certain knowledge bases (DBpedia, Geonames, Viaf and MusicBrainz) with the goal of:
- - helping the annotator in his documentation task by offering him rich, reliable information from web knowledge bases;
- - annotating the audiovisual content with individuals or concepts available on the web of data to ensure the interoperability of the local knowledge base with external bases;
- - connecting the local audiovisual content to external media available on the Web to improve its visibility and accessibility.