| ID |
Question |
Name |
| Giovanni Numeric Reasoning GTW1 |
The classic criticism of "Type 2 OBDBs" (and similarly structured plain triplestores) using a binary representation (vertical partitioning) is the inability to efficiently answer queries with unbound predicates. Similar problems exist with the mentioned "table per class representations" for storing multiple values per resource.
Given that the above mentioned storage schemes are the only ones discussed in the implementation of the proposed annotation and query rewriting mechanisms, could such a system be made to work with a more traditional triples-table storage scheme?
Bounding boxes are inadequate to test geographic subdivisions in the real world (the case of Baarle-Hertog, Belgium being perhaps the most notable and extreme, in which Dutch enclaves exist inside Belgian enclaves which exist inside of the Netherlands, proper). Can this system be made to work with more complicated geographic data than simple bounding boxes/circles, or is it constrained to the domain of simplistic examples? |
Gregory Todd Williams |
| Giovanni Numeric Reasoning Jesse Weaver |
The end of section 2.1 (under Type 3 OBDBs) states that "meta-schema may allow ... storage of different ontology models (OWL, DAML+OIL, PLIB, etc.)." Referring to figure 3 (as directed by the author) and to figure 7 (OntoDB, a type 3 OBDB), I find it difficult to perceive how such an approach would model OWL ontologies. In particular, it is unclear to me whether the Class table is able to take into account multiple inheritance of classes, and furthermore, there seems to be no account for subproperties. The paper focuses on numeric reasoning related to geospatial and temporal applications and doesn't seem to investigate the aforementioned claim. Do you think the meta-schema approach is capable of storing more expressive ontologies (like OWL ontologies)? |
Jesse Weaver |
| Giovanni Numeric Reasoning Joshua Shinavier 1 |
It makes sense that the technique presented could cut down on table joins in many cases. However, the mapping from "special" triple patterns (e.g. in a SPARQL query otherwise requiring logical reasoning) to PostgreSQL is not discussed in detail. Is it straightforward? |
Joshua Shinavier |
