Game Development Reference
In-Depth Information
which for the user represents the speed of the bear and makes perfect sense since
he is collecting articles to create ranking of animal speed. But from the univer-
sal point of view, this tag represents only a marginal part of the article. Another
drawback of user tagging is the relative generality of the used words (causing
overload of the tags lowering their distinctive ability) and often use of sentiment
or even irony with no semantic feedback on the content of the URL-identified
resources (e.g., “funny”, “gorgeous”, “stunning”) [
49
].
2.
Generate tag collocations
. When a user decorates a resource with more than one
tag, he expresses the, yet unspecified, relatedness of these tags. These collocations
are afterward used to create a lightweight tag folksonomy [
49
].
Even with all drawbacks, the user tagging is a potent source of lightweight semantics
and annotations. Using the agreement principle and dictionary, more clean metadata
can be obtained and utilized like, for instance, in the project
Treelicious
which com-
bines the delicious folksonomy with
Wo rd N e t
hierarchy to create a navigable tag
structure.
The associations between term relationships in folksonomies like delicious are
unknown, meaning that we do not know, what does the relationship means. Some
researchers build upon the folksonomies like Delicious and leverage their term rela-
tionships by identifying their specific types. In their work, Barla and Bieliková [
7
]
extract hierarchical relationship between terms with syntactical analysis of the term
graph structure.
2.5.4 Wikipedia and DBpedia
The DBpedia is an example of semantic building approach that mixes all manual,
crowdsourcing and automated approaches. It is closely related to world largest and
collaboratively created encyclopedia, the Wikipedia. Although the Wikipedia's arti-
cles can be seen as being created manually by pseudo-experts, the large number and
the collaboration of the contributors pushes it rather to the crowdsourcing category
of knowledge acquisition. While Wikipedia is primarily made readable to human
users, the DBpedia transcripts the knowledge contained in Wikipedia to a more
“machine-friendly” ontology, using RDF and OWL standards.
The basis of the DBpedia content is created automatically: Each article of
Wikipedia becomes a concept in DBpedia. Using various algorithms, the texts and
links of Wikipedia are mined to create relationships and assign properties to the
concepts. These include [
12
]:
Extraction of labels from titles and link named entities.
Abstract extraction from original article texts.
Article categories.
Geo-coordinates.
Properties through infoboxes. Infoboxes are structured information attached to
some articles, consisting of properties and their values. The infoboxes of articles
Search WWH ::
Custom Search