Spatial and Temporal Semantic Web

Linked Open Spatial and Temporal Data should not be… LOST Data

The context of our work in this axis is drawn by the now well-established paradigm of the Web of Knowledge that succeeds the Web of Documents. In this context, the key challenge is to enrich Web resources with semantics. That involves several tasks like categorizing these resources, indexing, linking them, or formally modeling the knowledge covered by these resources, with a final objective in mind: to allow software agents to extract, combine and eventually deduce information from published datasets. Ontology-based approaches, together with RDF graphs and SPARQL, offer a formal and explicit specification framework for shared conceptualizations about more or less specific domains of application, and also for reasoning on them.

We study how semantic technologies can foster data integration and interoperability when Spatio-Temporal (ST) concepts are involved, especially focusing on how to enhance knowledge from Linked Open Data (LOD). Regarding Geographic Information (GI), large crowdsourced datasets emerge from the Volunteered Geographic Information (VGI) phenomenon through platforms such as Open-StreetMap (OSM). This deluge of VGI consists of spatial features that are described with tags for content categorization. However, this approach is also a major impediment to interoperability with other systems that could benefit from this huge amount of bottom-up data, as folksonomies are much less expressive data models than ontologies.

In A. Hombiat’s PhD, we have addressed the issue of loose OSM metadata by proposing a meta-model for semantically lifting the OSM folksonomy while preserving the flexibility of the tagging activity. This meta-model supports the identification of di↵erent types of OSM tags and their semantification into the ontology OF4OSM. Comparatively to existing approaches, this ontology features a higher coverage of the OSM tags, an enhanced formal expressivity, a wider interconnection to other knowledge bases and a full compliance to the participatory philosophy of the OSM project.

This work paves the way to the building of a knowledge base of VGI structured data that is integrated in the Linked Open Data cloud. We have also particularly investigated how semantic technologies can assist researchers by matching data to models and different conceptual schemas between them and, lastly, how semantics-enabled user interfaces can support experts in exploring interdisciplinary datasets by browsing through interlinked data, and provide richer information through the discovery and exploitation of new connections between various resources.

This led us to develop and feed several ontologies: in the field of environmental sciences (water chemistry and hydrometric in the COIN project funded by the Natural Resource Canada (NRCan)); at the intersection between natural sciences, geosciences, engineering sciences and human and social sciences in the Trajectories ( or Patrimalp ( CDP projects; for some industrial partners such as the SNCF; or to contribute to knowledge dissemination for public bodies.

We have also developed specific algorithms and tools for helping experts of an application domain to manage and exploit semantics embedded in ontologies, as for instance for experts in geolinguistics in the ANR ECLATS project or for sociologists in D. Noël’s PhD about Life Trajectory. As another example, in the H2020 MICA project (, we have developed tools for the management and the efficient exploitation of an ontology in the mineral field that integrates heterogeneous resources (geographically dispersed, from both private and public organizations, following standards and regulations according to European directives but at the same time adapted to each country and legislation). The ontology identifies concepts, their relations, norms and standards used to describe mineral resources, and makes the national databases interoperable. It covers the areas of mining and mineral resource economics for the 27 mining and geological survey partners of the project, disseminating indispensable knowledge for many sectors that have a strong link with the mineral intelligence (the defense, economy, protection of the environment, development aid, land use planning, research and trade, etc.).

Beyond interoperability issues, several of our projects were characterized by the necessity to address knowledge evolution matters, both for a better understanding and for a better use of information. As an example of our works, C. Bernard’s PhD led to the definition of two innovative ontologies for the representation of any evolutive territorial partition. The Territorial Statistical Nomenclature (TSN) and TSN-Change enable richer descriptions of the successive versions of territories and of changes between them, than ever allowed in the literature. The TSN Semantic Matching Algorithm has been developed to manage the evolution of such TSNs: it automatically outputs Semantic Graphs that constitute catalogs of evolving statistical areas. The program has been implemented and tested on three different TSNs and their versions, exemplifying the genericity of our solution: European Nomenclature of Territorial Units for Statistics (NUTS) from Eurostat, the Switzerland Administrative Units (SAU) from The Swiss Federal Statistical Office, the Australian territory from the Australian Statistical Geography Standard (ASGS). The RDF graphs created (available as LOD at the URIs, and constitute catalogs of territorial units that draw the lineage of each over time (horizontal reading of the graphs) and the propagation of a change event through the divisions levels (vertical reading). This knowledge is exploited to extract patterns of change and simulate scenarios of territorial reorganization.