Information Technology Reference
In-Depth Information
•
A
synonym
allows the insertion of a word or a list of words that present the
same meaning as the class name. The same relevance weight in the ontology
for the class name is adopted for its synonyms during relevance calculations.
•
A
related verb
offers a similar meaning of a synonym except that the devel-
oped tool would try to obtain radicals for the listed verbs [24], which allow a
broader search for terms that represent one of the different verb conjugations.
•
A
relevance adjustment
is a specified value (0≤
v≤1
) that substitutes the de-
fault value that considers the hierarchy depth of a given class.
3.3 Object Properties
An object property allows interrelating classes or instances. In the relevance algo-
rithm for classes, the existence of such property is interpreted as the existence of a
cohesion grade between class families, independently of which class in the family
is involved in the property. An object property specifies that it is really relevant if
terms of both involved families are present in a given text.
It is possible to have an inverse object property, i.e., the existence of an object
property from an origin to a destiny together with another from the previous des-
tiny to the previous origin. For the retrieval algorithm this does not result in a
double relevance grade, just one.
4 The Information Retrieval Algorithm
The classic vector model was chosen because it provides a good computational
processing level combined to satisfactory results for most of known retrieval solu-
tions. Even though, modifications were done to adequate it to the desired semantic
contextualization.
In the following subsections are described the implemented modifications to
the original model together with their motivations and the mathematic impacts that
they produce.
4.1 Multiple Query Vectors
Each of the documents in a collection has its own term vector for which the dis-
tance is calculated regarding the term vector of a query or of an example docu-
ment, which is the base for retrieval. For the semantic approach, each class family
in an ontology gives origin to a term vector,
i.e.
, the ontology can produce several
vectors.
In addition regarding an ontology, a term equivalent is a set with the class or
instance name, and all synonyms and related verbs that are present in the annota-
tion properties.
While, in the original model, term weights are equal to the number of occur-
rences of a given term in one of the documents to be searched, in the semantic ap-
proach each occurrence of the name, or of each synonym or of each related verb
contributes with one to the weight in the term vector of a wiki element.
