State-of-the-Art: Semantics Acquisition and Crowdsourcing - Semantic Acquisition Games: Harnessing Manpower for Creating Semantics - page 12

2.2.3 Creation is the Harder Part

Though we have not covered other fields where semantics are used like recommender
systems [ 34 ] or learning frameworks [ 8 ], we have illustrated the important role of
semantics in today's Web applications. Many of them still await the “critical mass”
of existing web semantics in sufficient quality. We have seen that they do not need
a particularly rich (heavyweight) semantics to work. Nevertheless, even the light-
weight semantics is not available in sufficient scale.
Regardless on how the semantics are used or which type of resource they describe,
their creation usually represent the harder part in the job of semantics-based appli-
cations development. Many research works (including this) are devoted solely to
this task, leaving the semantics use to others. We now continue in description of the
existing approaches to semantics creation, which can be split into three categories:
manual, automated and crowd-based.

2.3 Manual (Expert-Based) Approaches

Manual approaches for building semantics rely on individual (or small groups of)
experts, who create domain models or resource descriptions. Because of the dedi-
cated human work, controlled environment and expertise the experts produce high
quality semantics. Their capabilities are, however, limited quantitatively. Employing
an expert in a specialized domain can be very costly. Additionally, experts need to be
trained to understand the concept, representation and tools for semantics definition
(in this task, they are sometimes aided with software tools [ 28 ]).
Expert work is essential for certain types of semantics acquisition tasks. In ontol-
ogy engineering, experts are needed for correct definition of the top layers of the
concept (class) hierarchy and system of ontology predicates and constraints. Another
example is creation of gold standard datasets or “grand truths” used in evaluation of
other semantics acquisition approaches.
The ontology engineering represents a field of study of creating ontologies and
covers the spectrum of manual (expert) domain model creation approaches. It covers
a variety of methods and methodologies comprising (not only):

￿ Strategies for defining relevant domain concepts [ 30 ].

￿ Methodologies for definition of relationship schema and ontology axioms [ 20 ]—

a step which greatly influences the expressiveness of the ontology and the capa-
bilities of automated inference over the ontology.

￿ Specialized methods for improving collaboration among experts during the ontol-

ogy building [ 43 , 48 ].

￿ What strategies to use in ontologymapping (i.e. interlinkingmultiple ontologies)—

a task which mostly involves seeking for equivalent concepts within different
ontologies [ 32 ].