Database Reference
In-Depth Information
Table 9. Q
1
results
Id_Sum
Price
Size
Location
z
1
cheapreasonableexpensive
mediumlarge
suburb
Price and consequently a new broad query with less selective conditions may be submitted or the
task may be abandoned (we denote this with the IGNORE option).
2:
summary
z
exactly fits the user's need. It means that for each
AY
, all linguistic labels in
z.A
are
relevant to the user. Assume that the user is interested in
cheap
,
reasonable
as well as
expensive
houses. Thus, all tuples contained in
R
z
1
are relevant to her/him. In such cases, she/he uses SHOW-
TUPLES option to access tuples stored in
R
z
1
.
3:
summary
z
partially fits the user's need. In this case, there is at least one attribute
AY
for which
z.A
exhibits too many linguistic labels w.r.t. the user's requirement. For instance, the set
R
z
1
partially
matches the needs of a user who is looking for
cheap
as well as
reasonable
houses because
R
z
1
contains also tuples that are mapped to
expensive
on attribute Price. In this case, a new query with
more selective conditions (e.g., Price IN {
cheap
OR
reasonable
}) may be submitted or a new
clustering schema of the set
R
z
, i.e., which allows to examine more precisely the dataset, is re-
quired. Since
z
is a subtree of the summary hierarchy, we present to the user the children of
z
(SHOWCAT option). Each child of
z
represents only a portion of tuples in
R
z
and gives a more
precise representation of the tuples it contains. For example, {
z
10
,
z
11
} is a partitioning of
R
z
1
into
two subsets
R
z
10
and
R
z
11
;
z
10
exactly fits user needs. Since the entire tree is pre-computed, no
clustering at all would have to be performed at feedback time.
More generally, a set of summaries or clusters
S
= {
z
1
…
z
m
} is presented to the user as a clustering
schema of the query result
tset(Q)
. The three options IGNORE (case
1
), SHOWTUPLES (case
2
) and
SHOWCAT (case
3
) give the user the ability to browse through the
S
structure (generally a set of
rooted subtrees), exploring different datasets in the query results and looking for potentially interesting
pieces of information. Indeed, the user may navigate through
S
using the basic exploration model given
below:
i. start the exploration by examining the intensional description of
z
i
S
(initially
i
= 0);
ii. if case 1, ignore
z
i
and examine the next cluster in
S
, i.e.,
z
i+1
;
iii. if case 2, navigate through tuples of
R
z
1
to extract every relevant tuple and thereafter, go ahead
and examine
z
i+1
;
iv. if case 3, navigate through children of
z
i
, i.e., repeat from step (i) with
S
, the set of children of
z
i
.
More precisely, examine the intensional description of each child of
z
i
starting from the first one
and recursively decide to ignore it or examine it (SHOWTUPLES to extract relevant tuples or