RDFS and OWL Reasoning for Linked Data - Reasoning Web - page 117

and XSD datatypes as used to type literals in our corpus. We omit plain literals
which were used in 6.609 million documents (89%). The

Lit

column indicates
the number of literals with that datatype. D indicates the datatypes supported
by D-entailment with the recommended XSD datatype map. O2 indicates the
datatypes supported by OWL 2.
We make the following observations based on Table 6:

1. The top four standard datatypes are supported by both the traditional XSD
datatype map and by OWL 2.
2. OWL 2 does not support xsd:date (5) , xsd:time (26) , or the various Grego-
rian datatypes (10,15,18,20,22) .
3. Despite not being supported by any standard entailment, xsd:duration (14)
was used in 28 thousand documents across four different domains.
4. Various standard datatypes are not used at all in the data. For example,
xsd:dateTimeStamp , the “new” OWL datatypes, binary datatypes and var-
ious normalised-string/token datatypes did not appear at all. 22

4.4 A Profile of OWL for Linked Data?

Our analysis of the adoption of RDFS and OWL has shown that while some
features are broadly adopted on the Web of Data, others are barely adopted at
all. Thus, it would seem possible, for example, to only support some fraction
of the standard OWL features while capturing support for the broad majority
of axioms present on the Web of Data. In Table 5, we already saw that the
most frequently used features corresponds with the ability to represent their
respective axioms in RDF as a single triple (and thus without blank nodes if
being interpreted under the Direct Semantics).
In previous work, we thus proposed the OWL LD (Linked Data) profile, which
is a proper subset of OWL 2 RL supporting only those features that are express-
ible with a single RDF triple [22]. The RDF-Based Semantics of the OWL LD
profile can be (partly) supported by means of a subset of the OWL 2 RL/RDF
rules relating to the supported features. We also provide a grammar under which
the Direct Semantics of the profile can be supported, making (optionally) confor-
mant documents compatible with the OWL Direct Semantics. We propose that
OWL LD - as a practical, terse profile - facilitates greater ease of implementa-
tion for Linked Data reasoning applications, while maintaining high coverage of
commonly used features.

5 Rule-Based Inference for Linked Data: Authoritative
vs. Context-Dependent Reasoning

A common strategy for reasoning over multiple sources is to simply merge them
together and compute the deductive closure over the resulting monolithic RDF

22 In fact, owl:real does not have a lexical space, and thus cannot be written down;
irrational numbers are di cult to write down.