Queries and Answers in Concept-Based Knowledge Bases
Jun 1998
This thesis aims at closing the gap
between the expressive power of terminological languages,
in which knowledge bases are written,
and the expressive power of the language of answers.
To this end, a new framework for dealing with elements
that have no explicit representation in the knowledge base
is proposed.
With an appropriate representation for those elements,
answers are possibly infinite regular languages,
which are described by regular expressions.
Hence, in contrast to the other formalisms,
answers are not restricted to be a finite set of individual constants,
and
the query-answering algorithm does not work by generate-and-test.
The declarative semantics extends
the classical definition of answer in a very natural way.
The operational semantics relies on a finite automaton
which is constructed, independently of the queries,
from some completions of the knowledge base.
Because this approach turned out not to suit every case,
knowledge bases are classified
into two groups --- cyclic and cycle-free ---
according to the relational assertions they possess.
Problems stem from cyclic knowledge bases,
for which there may not be a finite automaton that
captures the information needed to answer queries.
For this reason, the operational semantics,
which is sound and complete
with respect to the declarative semantics,
is defined for cycle-free knowledge bases.