IMADA - Department of Mathematics and Computer Science |
The first part of this talk focuses on relational keyword search
(R-KWS). Under this paradigm, users access a database through simple
Google-style queries. Compared to SQL, it has several benefits. First,
R-KWS hides the schema from users and allows searching for combinations
of terms without knowing in which data sources they appear. Second, it
queries are easy to express. Third, many search tasks only become
feasible through R-KWS. There are two general methodologies for query
processing. Graph based (GB) systems model the database as a graph,
and retrieve results by means of traversal. In contrast, (ii) operator
based (OB) systems solve a query by executing a (large) set of
relational operator trees. Both methods consume considerable resources,
due to the vast search space, composed of all possible combinations of
keyword occurrences in any attribute of every table. This talk presents
homogenized R-KWS semantics, and basic query processing methods (GB as
well as OB). Next, it introduces a comprehensive framework for indexing
reachability along join-sequences that quickly reduces the search space,
and greatly accelerates query processing. Subsequently, it extends R-KWS
to continuous queries over relational data streams. Finally, we learn
about searching for "relational outliers" and other oddities.
SDU HOME | IMADA HOME | Previous Page Daniel Merkle |