Conclusions - Ranking Queries on Uncertain Data - page 210

9.2.1 Top-k typicality queries for uncertain object

In this topic, we study three types of top- k typicality queries on static uncertain ob-
jects. As future study, it is interesting to extend this study in two directions. First,
how can we answer top- k typicality queries on different types of uncertain data,
such as uncertain data streams, that is, uncertain objects with evolving distributions?
Second, how can we develop alternative of typicality notions that fit different appli-
cation needs? For example, when the dimensionality of uncertain objects is high,
finding the most typical instances in the full space may not be informative. Instead,
it is interesting to find in which subspace an instance is typical.

9.2.2 Top-k queries on probabilistic databases

Chapter 5 discusses the problem of top- k query evaluation on probabilistic databases.
It is interesting to investigate the following two extensions.
First, the probabilistic database model we adopted in Chapter 5 only considers the
generation rules that specify the mutual exclusiveness among tuples. More complex
generation rules can be considered. For example, mutual inclusion rules that specify
the coexistence of the tuples involved in the same rule are discussed in [7]. A mutual
inclusive rule R ≡ : t r 1 ≡···≡

t r m involved
in the same rule, either no tuple appears or all tuples appear in a possible world.
All tuples in R ≡ have the same membership probability value, which is also the
probability of rule R ≡ . How to answer top- k queries on probabilistic databases with
various generations rules is an interesting extension.
Second, how to incrementally update top- k query results when changes happen
to probabilistic databases? For example, the membership probability of an uncertain
tuple may be updated when more knowledge about the data is obtained. Instead
of recomputing the results based on the updated probabilistic database, it is more
efficient to reuse the results computed before updates happen.

t r m restricts that, among all tuples t r 1 ,···,

9.2.3 Probabilistic ranking queries

In this topic, we study probabilistic ranking queries on probabilistic databases, un-
certain streams, and probabilistic linkages. It is interesting to extend our methods
to other probabilistic data models and other ranking queries. Particularly, there are
four important directions.