Game Development Reference
In-Depth Information
How to clean existing ontologies. Realizing repetitive problems and mistakes
occurring during the ontology building, some researchers came up with sets of
guidelines for clearing inconsistencies within ontologies [
27
].
The physical creation of the ontology is then done through software tools like
Protégé
1
(WebProtégé [
66
]) or HOZO [
47
].
Experts are also employed in resource description acquisition tasks, mostly for
commercial purposes. Typical examples are photography, 2D textures or press agency
databases. Usually, such corpora are not freely available to public (paid services for
professionals, e.g., journalists, designers) and do not follow the description standards
of the Semantic Web (i.e. using standards like RDF and linking to global ontologies).
However, the principles that are used there are similar to those on of the Semantic
Web and such corpora could be eventually (straightforwardly) transformed to meet
the Semantic Web standards.
Quantity of delivered semantics is the major disadvantage of expert-based seman-
tics acquisition approaches. However, even manually created knowledge base can
grow in size, if it is given enough time and effort. In the
Cyc
project,
2
a general knowl-
edge base has been constructed for 25 years [
38
]. The project is being developed for
commercial purposes (to be utilized, for example, by expert systems), but it has also
been made partially published through the
OpenCyc
release—which demonstrated
the impressive scale of 47 thousands of concepts and 306 thousands of facts (triplets
or property assignments). Other positive aspect of the Cyc is its exhaustiveness: it
covers—at least on top levels of abstraction—the whole spectrum of human knowl-
edge. Another advantage is the inclusion of common sense facts that are not present
in a written form, simply because they are commonly known (e.g., “You cannot
remember events that have not happened yet.”).
The Cyc knowledge base is heavyweight: it is highly structured and provides
also a reasoning engine to answer even complex logical questions. This, however,
comes with a price: the eventual volunteer effort to contribute new facts or even the
usage of the knowledge base becomes a difficult task due to its complexity. The Cyc
critics also note numerous gaps in the ontology: while there is enough concepts the
relevant relationships among them often miss [
72
]. But even with these drawbacks,
the Cyc knowledge base is usable and expandable and can be a good benchmark for
evaluation of automated semantics acquisition approaches.
Another good example of expertly created knowledge base is the
Wo rd N e t
3
dictio-
nary [
24
], used by many web semantics projects. Created and maintained at Prince-
ton university since 1985, it contains English language words organized by
synsets
(according to their synonymic relatedness), parts of speech, lexemes (various tex-
tual forms of the same term) and other relationships (e.g., hypernyms, holonyms).
In comparison to Cyc, WordNet is a lightweight corpus. It operates over words, not
concepts. Its set of relationships between words is limited and very abstract. Axioms,
Search WWH ::
Custom Search