RDFS and OWL Reasoning for Linked Data - Reasoning Web - page 108

and store the closure lm ( merge ( Γ )

R ), 15 , issuing queries as they are input di-
rectly against the closure. Caching the full closure thus mitigates the expense of
reasoning during query-time, but may conversely incur huge pre-runtime costs,
storage overheads, as well as the cost of truth maintenance in dynamic scenarios.

∪

Example 4
As an example, let us consider the following variant of Query 3: (i) in-
stead of explicitly following the owl:sameAs link, we assume the necessary
inferences are supplied by reasoning; (ii) we ask for all rdfs:label s ofthe
company (as opposed to just skos:prefLabel ) .

Query 3
SELECT ?D ?R ?L
WHERE { nyt:49586210195898795812 nytimes:latest_use ?D ;
dbo:revenueUSD ?R ;
rdfs:label ?L }

while the first triple pattern is matched by explicitly stated data, the subse-
quent query-relevant triples must be obtained from the closure with respect
to, e.g., the rules in Table 3, which contains (amongst other triples):

nyt:49586210195898795812 nytimes:latest_use "2010-04-27"^^xsd:date ;
dbo:revenueUSD 1.06916E11 ;
rdfs:label "IBM"@en ,
"International Business Machines Corporation"@en .

leading to the following solutions:

?D ?R ?L
2010-04-27 1.06916E11 "IBM"@en
2010-04-27 1.06916E11 "International Business Machines Corporation"@en

♦

3 Overall Approaches and Challenges to Reason over and
Query Linked Data

We identify two main approaches to reason and query over Linked Data:

1. Data-warehousing approaches for querying linked data are typically deployed
for RDF search engines such as Sindice [54] or SWSE [40]. These engines pro-
vide query interfaces over a local centralised index of Linked Data harvested

15 That is, all RDF triples entailed by the RDF graph obtained from Γ (read as facts)
and R .