December 16th, 2021
A declarative approach based on Semantic Web technologies to specify and generate adaptive geovisualisations
Abstract
The Semantic Web technologies proposed by the W3C have demonstrated their strength when it comes to exchanging data (with RDF), formalising various knowledge domains (through RDFS/OWL ontologies) or querying this knowledge (with SPARQL). Thus, today, many models make use of Semantic Web technologies and describe a domain with a geospatial dimension possibly relying here on specific models to describe this spatial dimension at a high level, e.g. GeoSPARQL.
In GI Science, besides knowledge modeling and representation, the main challenge remains that of the geovisual restitution of the information. Indeed, geospatial information and the interfaces for viewing and interacting with it, commonly referred to as geovisualisation applications, play an indispensable role in the understanding of various spatial phenomena and in the decision-making processes involving this information.
This works explores how Semantic Web technologies can facilitate the specification and generation of adaptive geovisualisations. Indeed, classical approaches for creating geovisualisation from RDF data involve numerous transformations of the data, which can lead to the loss of the semantics encoded in the models. In addition, approaches allowing the direct geovisual exploitation of RDF data described by ontological models are rare and have significant drawbacks. Besides, making cartographic portrayals and geovisualisation applications that make sense is a difficult task that is worth assisting, especially when the designer is not a cartographer.
In this thesis we present the CoViKoa framework. CoViKoa allows to describe, through a purely declarative specification document, how to create a geovisualisation for an RDF model and dataset. To do this, CoViKoa adopts Semantic Web technologies.
First, it is based on an ecosystem of ten ontologies, six that we propose and four that we reuse from the literature. Secondly, it is instrumented by SHACL rules, derived from the specification document, which allow to establish the links between data and the way they appear in the components of a geovisualisation. Finally, the proposed framework allows the publication of the resulting RDF graph behind a SPARQL interface. This RDF graph contains all the necessary elements for building a geovisualisation application. It is exploited by a Web application that we propose and that allows to concretely build the corresponding geovisualisation. The geovisualisations that can be described with CoViKoa are rich and complex: various mechanisms of data transformations, filters and selections are proposed to the user, directly in RDF. It is also possible to describe several types of interactions between component data. The CoViKoa framework is mainly intended for a knowledge engineer who would like to geovisually leverage the geospatial RDF data he or she has. Nevertheless, since it is based on rich ontologies that can describe many aspects of geovisualisation applications, it is also intended for geovisualisation researchers who can use these vocabularies as an analytical framework to describe and compare geovisualisation applications while benefiting from the power and expressiveness of the SPARQL query language to conduct their analyses. We also validate our proposal with two case studies. The first one focuses on the processing of the search for a lost person in mountain environnements, and the second one deals with the spatio-temporal evolutions of statistical territorial entitie
ARQL pour mener leurs analyses. Nous validons également notre proposition par deux études de cas. La première porte sur le traitement de la recherche d'une personne perdue en milieu montagneux, et la seconde traite des évolutions spatio-temporelles d'entités territoriales statistiques.
Mis à jour le 9 December 2021