Spatial queries and inference
The early attempts to store geographic data in standard relational tables failed. This issue gave rise to spatial databases, which extent the relational model providing specific representations for geospatial data.
As the number of geographic features in a database grew, it became evident that it was imposible to store all the relationships between geographic entities in a large GIS. So distances, intersections, centroids, etc. are easyly computed with built-in functions. Questions like the depicted in the image could be seen as real-time spatial inferencing.
Is featureA within a 5km distance from the border of any water body feature?
When developing spatial reasoning support for Semantic Web we find the same problems. Description logic standards as OWL don’t provide specific spatial built-ins to enable real-time spatial inferencing or querying.
So we need the preprocessing of all spatial-related statements using spatial specific algorithms (or GIS) before adding them to the semantic KB or we need the adding of some kind of spatial layer that combines spatial queries with semantic queries to extent DL expressivity.
We use both approaches in our spatial model for CINeSPACE project:
- A preprocessing of fuzzy spatial relationships such as mereology or neighbourhood, that are added to the KB. This is an feasible approach while the number of spatial elements remains low (hundreds of elements for each city) and the spatial extent of polygons remains constant (non-monotonicity will be reviewed in a future post).
- A spatial layer used for dynamic queries, such as “Retrieve me Geoconcepts I am looking at”, where geoconcept is used to name any entity with an inherently or indirectly associated spatial dimension. SPARQL queries are used over the spatial results to refine the answer.
The need of extra spatial computation limits the success of Geospatial Semantic Web in terms of GIS professionals, who see Semantic Web as a complex world with not great benefits for dealing with space. A step forward spatial integration with DL logics. We are looking for researchers and people interested in this integration. Feel free to contact!!!
