Database Reference
In-Depth Information
Figure 1. Lattice of generalized queries of q (cf. Motro, 1986)
conditions in the query. As an example (cf. Motro, 1986), consider the following query q over the Em-
ployees' relation EmpDB:
SELECT *
FROM EmpDB
Q ≡ WHERE Age ≤ 30
AND Gender = F
AND Salary ≥ 40K€.
Figure 1 shows a portion of the lattice of relaxations of q generated by SEAVE, where nodes indicate
generalizations (or presuppositions) and arcs indicate generalization relationships. ( x , y , z ) denotes a
query which returns the employees whose age is under x , whose sex is y , and whose yearly salary is at
least z . The symbol * indicates any value; once it appears in a query, this query cannot be generalized
any further. Assume in this example that q fails and all its relaxed queries are successful queries except
those that are marked √ in Figure 1 (i.e., q 1 , q 3 and q 8 ). Thus, the failed relaxed queries q 1 and q 3 are
XGFs, whereas the successful relaxed queries q 2 , q 4 , q 6 and q 15 are MGSs. The queries q 2 , q 4 , q 6 and q 15
produce alternative answers to the original query q : ( q 2 ) all employees under 30 who earn at least 40K€;
( q 4 ) all female employees under 32 who earn at least 40K€; ( q 6 ) all female employees under 31 who
earn at least 39K€; and ( q 15 ) all female employees under 30 who earn at least 37K€. These answers can
be delivered by SEAVE to the user as 'the best it could do' to satisfy the query q .
The main drawback of SEAVE is its high computational cost, which comes from computing and
testing a large number of generalizations (i.e., various combinations of the values of attributes) to iden-
tify MGSs/XGFs.
CoBase
The CoBase iii (Chu, Yang, Chiang, Minock, Chow & Larson, 1996) system augments the database with
Type Abstraction Hierarchies (TAHs) to control the query generalization process. A TAH represents at-
tribute values at different levels of granularity. The higher levels of the hierarchy provide a more abstract
data representation than the lower levels (or attribute values). Figure 2 shows an example of TAH for
Search WWH ::




Custom Search