Database Reference
In-Depth Information
note that the analysis of the resulting social network also yields interesting global
results.
A somewhat similar approach is described in [ 22 ]. There, a central repository is
created using a generic schema, with domain experts expected to simply take care
of uploading data into the system using a simple, general RDF interface. The data
and the schema of the database are customized as much as possible, using the fact
that from all declarations from the experts, some patterns can be observed that
allow grouping and organization of data.
The paper [ 23 ] proposes a system for building content sharing communities .
This system allows users to share their data. This is similar to our idea of allowing
users to share their private data. However, [ 23 ] does not presuppose an existing
database, and therefore it is free to choose a novel structure for the shared data and
its querying (based on Active XML over data streams). In our framework, a
preexisting (relational) database is taken as the starting point. Nevertheless, there
are interesting similar aims between the proposal of [ 23 ] and ours.
Another approach [ 6 ] also proposes to capture user data in a flexible format and
allow sharing in a central repository. As in our proposal, a generic relational table is
used to hold user-created content. Here, a flexible schema in which attributes are
added on the fly is proposed. The data is then clustered by common attributes in
an attempt to define the types of objects that the users are referring to. Note that
this schema leads to numerous null values, since not all objects will have values
for all attributes. By contrast, our approach stores only needed information. In [ 6 ],
each type of object leads to the creation of a view. A collection of such views is
offered to users for querying. This work, like [ 23 ], does not assume a preexisting
database. Hence, the data from users must be analyzed to identify the entities under
consideration. We anchor all user-created content in the database. In other words,
we focus on parasitic data, while these efforts focus on independent data (see
Sect. 7.3).
The papers [ 24 ] and [ 25 ] also develop systems to store and query annotations.
Geerts et al. [ 24 ] presents the MONDRIAN system, based on a special algebra to
handle together annotations and the tuples they refer to. The resulting language
gives a “parallel” relational algebra to manage annotations. While this is an
interesting approach, it presents serious challenges from an implementation point
of view. Here we have chosen to manipulate all information in traditional relational
algebra in order to make the resulting system as simple as possible. Bhagwat et al.
[ 25 ] introduces the DBNotes system. This system focuses on annotation propaga-
tion : as data is transformed, the associated annotations are transparently carried
with the data. This is carried out using ideas from research on data lineage
(provenance). We have not covered this dynamic aspect here.
Recently, work on emergent semantics [ 26 ] has added emphasis to user-created
content, proposing it as a means to add semantics to different objects, like multi-
media objects, which are hard to define with traditional modeling [ 27 ]. Our system
can be seen as an enabling mechanism to capture emergent semantics in databases
and is partly motivated by this research, which considers user input as first-class
information, i.e., just as, or important than, the data stored in the system.
Search WWH ::




Custom Search