Information Technology Reference
In-Depth Information
The concept hierarchy illustrated as figure 1 possesses a nested and hierarchi-
cal structure, where every node represent a concept, and arcs represent partial
order relation between these concepts. Nodes at lower level are represent specific
concepts, and those at higher level are represent abstract concepts.
Different people may own different prior knowledge, and different people may
also have different preferences. Thus, for a given problem, different people will con-
struct different concept hierarchy. To illustrate this, we still take 'education' for
example. If you are a manager of some college and university, or a manager of a
science institute, your prior knowledge relevant to 'education' may be structured
as figure 1. But in common people's opinion, 'undergraduate' should also belong
to high education. Thus, prior knowledge is relevant to the context of a problem.
A concept hierarchy can only organizes relevant prior knowledge from one
particular angle or point of view with multiple levels of granularity, but it can
not organizes those from multiview.
In the next section, we will borrow the concept of multidimensional data
model from data warehousing, and provide an organization of prior knowledge.
This organization can represent relevant prior knowledge from multilevel and
multiview.
Multidimensional data model.
In 1969, Collins and Quillian [15] made a typ-
ical experiment to prove prior knowledge that stored in long-term memory are
in network architecture. This experiment suggests that people organize knowl-
edge structurally and stored the features of the concept in different levels of the
hierarchical architecture.
How can we represent this network architecture intuitively? In this subsec-
tion, we will organize multiple concept hierarchies as an organic whole, which is
represented by a multidimensional data model. This organization can not only
make relevant prior knowledge more easily understandable, but also represents
them from multiview and multilevel intuitively.
Multidimensional data model [20,21] is a variation of the relational model that
uses multidimensional structures to organize data and express the relationships
among data, in which the data is presented as a data cube, which is a lattice of
cuboid. A multidimensional data model includes a number of dimensions that
each includes multiple levels of abstraction defined by concept hierarchy. Thus,
a multidimensional data model can be treated as a combination of multiple
concept hierarchies, and it can represent data from multiview and multilevel.
This organization provides users with the flexibility to view data from different
perspective. Based on the hierarchical structure of multidimensional data model,
it is possible to “scan” a data table from different levels of abstraction and
different dimensions.
In a multidimensional data model, each dimension can be represented by
a concept hierarchy, which can represents a problem from a particular angle.
Concept hierarchies of multiple dimensions be organized as a multidimensional
data model, which can represents data from multiview and multilevel. In fact,
multidimensional data model itself is a well-organized network structure, it can
reflect relevant prior knowledge intuitively and completely.
