Summary Abox Ankesh Jesse Weaver
From Semantic Portal Wiki
- Question is for the Presentation: Summary Abox Ankesh
- Question is asked by: Jesse Weaver
- The Question is: Section 3.2 states: "We make the simple observation that if a role R and its inverse R− are not part of any universal or maximum cardinality restrictions, then R(a, b) can never be used to alter the original Abox or detect a clash, so it can be ignored." Looking at the LUBM ontology (http://www.lehigh.edu/~zhp2/2004/0401/univ-bench.owl), there are no universal or maximum cardinality restrictions. This being the case, it would seem that after filtering, LUBM summary data should be reduced to all single-individual partitions. However, according to Table 3a, that is not the case. Have I overlooked something?
- Answer: Thanks for pointing me to the results in the table. I don't think you have over-looked anything. At the same time I have no definitive answer. I would guess the following:
- 5 roles remain in the summary Aboxes. The paper says nothing about role-assertions when roles are of datatype. (Instinctive feeling is that they should be dropped too because we do not perform any reasoning on data. Moreover in SHIN you do not have functional or inverse-functional properties.) In all there are 7 datatype properties, of which when I checked univ0_0.owl there is no use of age, title and office-number. So 4 out of 7 datatype properties occur in extensional knowledge. That leaves 1 to be explained.
- paper doesn't discuss transitive properties in detail. So it appears that they do not filter out transitive roles under any condition. In LUBM subOrganizationOf is the only transitive property. This could explain the 5th property in summary Abox. Further on checking univ0_0 I could find two types of individuals- ResearchGroup and Departments, as subjects of this property.
- Then why 2 partitions? explanation 1- 1 partition contains all datatype properties. 2nd partition contains suborganizationof property extensions. explanation 2- both partitions contain extensions of subOrganizationOf. 1 contains individuals from researchGroup and other from Departments. (But none of the researchgroup or department instances are set to be different. and can not explain 5 roles in summary Abox).
Facts about Summary Abox Ankesh Jesse WeaverRDF feed
| Question answer | Thanks for pointing me to the results in t … Thanks for pointing me to the results in the table. I don't think you have over-looked anything. At the same time I have no definitive answer. I would guess the following:
|
| Question asked | Section 3.2 states: "We make the simple ob … Section 3.2 states: "We make the simple observation that if a role R and its inverse R− are not part of any universal or maximum cardinality restrictions, then R(a, b) can never be used to alter the original Abox or detect a clash, so it can be ignored." Looking at the LUBM ontology (http://www.lehigh.edu/~zhp2/2004/0401/univ-bench.owl), there are no universal or maximum cardinality restrictions. This being the case, it would seem that after filtering, LUBM summary data should be reduced to all single-individual partitions. However, according to Table 3a, that is not the case. Have I overlooked something? ot the case. Have I overlooked something? |
| Question asked by | Jesse Weaver + |
| Question for the Presentation | Summary Abox Ankesh + |

