This work is part of the PRIME TEATIME team. Heritage data are massive and stand out for their diversity and multidimensional nature. They simultaneously refer to space (3D), time (3D + T), as well as a multitude of thematic dimensions (nD), creating an exceptionally rich informational fabric. This thesis aims to address the challenges posed by the volume and heterogeneity of these data sets, and to significantly improve their exploitation by developing innovative navigation and aggregation methods.
It is possible to aggregate nD data, whether in 3D environments, or as a graph by using methods that allow identifying the various relationships that link them together (semantic, spatial, temporal, provenance, etc.). However, these terms often prove to be unintelligible for users. It is then necessary to find ways to simplify, and compress, the information, to offer navigation methods that make more sense for users according to their profiles and scientific questions.
The objective of this thesis is to develop methods for extracting simplified views from the data sets, guided by a certain form of dimensional proximity, by the scale of representation, or even by the interests of users. The proposed methodology includes an in-depth analysis of existing heritage data sets, the design and implementation of projection and compression algorithms to build levels of detail adapted to different dimensions of heritage data (X-LOD), as well as the creation of new navigation interfaces.