Database Reference
In-Depth Information
Figure 6. Example of a grid in two dimensions
9]
A
2
[0, 10]) are returned to that user.
C
1
contains 6 tuples
t
1
,
t
3
,
t
4
,
t
6
,
t
9
and
t
10
, whereas
C
2
contains 4
tuples
t
2
,
t
5
,
t
7
and
t
8
.
This approach is efficient. Indeed, it relies on a bucket-level clustering, which is much more efficient
than the tuple-level one, since the number of buckets is much smaller than the number of tuples. How-
ever, the proposed algorithm requires the user to specify the number of clusters
k
, which is difficult to
know in advance, but has a crucial impact on the clustering result. Further, this approach generates flat
clustering of query results and some clusters may contain a very large number of results, although, this
is exactly the kind of outcome this technique should avoid.
2.3 Flexible/User-Friendly Database Querying
A typical problem with traditional database query languages like SQL is a lack of flexibility. Indeed,
they are plagued by a fundamental problem of specificity (as we have seen in Sections2.1 and 2.2): if
the query is too specific (with respect to the dataset), the response is empty; if the query is too general,
the response is an avalanche. Hence, it is difficult to cast a query, balanced on this scale of specificity,
that returns a reasonable number of results. Furthermore, they expect users to know the schema of the
database they wish to access.
Recently, many flexible/user-friendly querying techniques have been proposed to overcome this
problem. The main objective of these techniques is to provide human-oriented interfaces which allow
for a more intelligent and human-consistent information retrieval and hence, diminish the risk of both
empty and many answers. Examples include preference queries (Section 2.3.1), fuzzy queries (Section
2.3.2) and keyword search (Section 2.3.3).
2.3.1 Preference Queries
The first way of introducing flexibility within the query processing is to cope with user preferences.
More precisely, the idea is to select database records with Boolean conditions ('hard' constraints) and
then to use preferences ('soft' constraints) to order the previously selected records.
Lacroix and Lavency (Lacroix & Lavency, 1987) were the first to introduce the notion of a preference
query to the database field. They proposed an extension of the relational calculus in which preferences
for tuples satisfying given logical conditions can be expressed. For instance, one could say: pick the
