Information Technology Reference
In-Depth Information
A
topic
in a wiki corresponds to a document. It is composed by an article and a
discussion. Regarding relevance, a calculation is done for the article and other for
the correspondent discussion. The higher relevance grade of both is considered to
be the topic
relevance.
In order to decide if topic part is relevant, the information retrieval vector
model is used. Considering a set of documents and a query to retrieve the most
relevant documents, each document is represented as a vector. Each vector ele-
ment corresponds to a separate term in the document set upon which the query is
performed. If a term occurs in a specific document, its value in the corresponding
vector is non-zero. There are different ways of computing term values, also known
as term weights. Considering
n
as the amount of terms in a vector of a specific
document set, each vector can be seen as a point in a
n
-dimensional space. Simi-
larly a vector is defined for the query, as it was a document. The similarity of one
document and the query can be measured by the distance of their correspondent
n
-
space points.
The information retrieval vector model is in essential a classification model that
allows handling large volumes of data [6].
The mathematic formulas of the classic vector model were modified in order to
consider the semantic nature of the ontology elements, such as the case of a con-
cept that is related to other concepts. The relevance of several associated semantic
terms should be weighted higher than the relevance of a isolated one.
3 Ontology Definition
A tool was designed to retrieve the recent and relevant content of a wiki, whose
location should be informed together with the ontology to be considered.
Using OWL,
Annotation
and
Object Properties
, synonyms, related verbs, and
relevance weighting adjustments can be incorporated into each of the ontology
classes in order to allow the calculation of relevance grades.
3.1 Classes and Instances
The Protégé editor allows the creation and maintenance of classes, subclasses and
instances in an ontology. Class names should be keywords that reference main
concepts in the domain of interest.
Relevance grades for a given class are calculated according to the depth of the
class in the ontology hierarchy. For instance, in a three level class hierarchy, a first
level class receives a 0.33 relevance weight. A second level class receives a 0.66
and the leaf classes in the hierarchy tree receive a 1.
The developed tool, when querying classes, considers composed words through
underscore identification or through the CamelBackCase syntax.
3.2 Annotation Properties
Three annotation properties were chosen to extend the semantic meaning of
classes or instances in an OWL ontology. Their meanings are:
