Data Normalization - Information Systems Reengineering, Integration and Normalization - page 347

Unnormalized form
1NF
2NF
3NF
BCNF
4NF
5NF
Fig. 9.10 Levels of relational data normalization
whenever such a data semantic in an XML document is found, we add appropriate
reference classes and classes to XML document to remove the data redundancy.
For example, if (V, W) denotes a many-to-many relationship in an XML document
where V is the parent node and W is the child node, then we add a new class W to
the XML document and replace W in (V, W) by (V, W^) where W^ is a reference
class to W. In other words, W^ denotes the IDs, or keys, of W. In this way, potential
data redundancy is removed.
Of all the data semantics, only four data semantics may cause data redundancy.
They are: overlap generalization, partial participation, m-to-n cardinality and n-ary
relationships. The arguments for the other data semantics of this chapter that do not
cause data redundancy are as follows:
Specialization: Specialization does not cause data redundancy because each
instance of a specialization is also an instance of the generalization. Hence, it does
not have a many-to-many relationship from the parent node to the child node.
Therefore, it does not cause data redundancy.
Disjoint Generalization: This case is similar to the case of specialization except
there is more than one specialization. Additionally, the specializations are all pair
wise disjoint. Thus, there is no data redundancy.
Categorization: In this case, there is only one specialization but there are mul-
tiple generalizations. However, an instance of a specialization is an instance of only
one generalization by definition. As such, there is no data redundancy.