Databases Reference
In-Depth Information
potential expense of such reasoning, and given that equivalence relations cannot
be universally trusted on the Web [29], a number of works have tackled this is-
sue with specialised techniques and optimisations [42,44,68]. For example, most
systems supporting
owl:sameAs
reasoning at large scale use a single canonical
identifier to represent each set of equivalent identifiers, avoiding the explosion
of data that could otherwise occur [39,42,68,11]. In previous work, we applied
authoritative reasoning to compute
owl:sameAs
relations from functional and
inverse-functional properties and cardinality restrictions [42].
Interestingly, Hu et al. [42] investigate a notion of authority for
owl:sameAs
inferencing, assigning a level of trust to such a relation based on whether the
given document is authoritative for the subject or object or both of a same-as
relation (here applying authority on an A-Box level). In any case, we note that
owl:sameAs
is an important reasoning feature in the context of Linked Data,
but similarly requires specialised techniques - that go beyond a generic reasoning
framework - to handle effectively in scalable, real-world settings.
6 Enriching Link-Traversal Based Querying of Linked
Data by Reasoning
As discussed previously, data-warehousing approaches - such as those introduced
in the previous section - are not well suited for reasoning and querying over
highly dynamic Linked Data. Content replicated in local indexes will quickly
become out-of-date with respect to the current version of the respective sources
on the Web. However, we referred in the introduction to the vision of the Web of
Data itself as being a giant database spanning the Web, where certain types of
queries can be posed and executed directly over the sources it contains. Such an
approach for executing SPARQL queries directly over the Web of Data - called
Link Traversal Based Query Execution (LTBQE) - was first proposed by Hartig
et al. [33] (
6.1). However, the original approach did not offer any reasoning
capabilities; indeed, no existing reasoning approaches at the time would seem
suitable for such a scenario.
In this section, we first describe the LTBQE algorithm (a comprehensive study
of the semantics and computability of LTBQE has been covered in [32]), complete
with formal definitions and illustrative examples, motivate why RDFS/OWL
reasoning is useful in such a setting, and then discuss methods we have ourselves
since proposed to support such reasoning features.
ยง
6.1 Overview of Baseline LTBQE
Given a SPARQL query, the core operation of LTBQE is to identify and retrieve
a focused set of query-relevant RDF documents from the Web of Data from
which answers can be extracted. The approach begins by dereferencing URIs
found in the query itself. The documents that are returned are parsed, and
triples matching patterns of the query are processed; the URIs in these triples
are also dereferenced to look for further information, and so forth. The process
is recursive up to a fixpoint wherein no new query-relevant sources are found.
