Abstract:
Ontology is used to express the concepts of domain
knowledge. It can provide a common representation for
different agents to share and communicate knowledge for
conducting unified opinions. Nowadays ontology construction
method is divided into man-made and machine-made
mechanisms. The former constructs the ontology topology by
domain expert. Generally the constructed ontology can fit
human expectation but it needs more development time to
construct the whole structure. The latter uses semi-automatic
or automatic methods, such as statistic or machine learning, to
build the ontology. The efficient ontology construction is the
main advantage for machine-made method. However, the
advantage is that it is easily influenced by the category and
type of domain concepts to generate unbalanced or skewed
ontology topology. This will increase the time complexity to
search and retrieve the concept from the constructed ontology
structure. The situation worsens from being unable to use the
ontology properly. An important problem is constructing a
reasonable and balanced ontology topology systematically and
automatically. This paper proposes a structured ontology
construction based on data clustering and pattern tree mining.
The construction method uses data clustering and formal
concept analysis to group similar documents for constructing
ontology trees of each group individually. Then the method
uses pattern tree mining to build an integrated ontology
topology from partial ontology trees.