|
|
|
The following
research questions were defined:
|
|
1) Define
an inference process on top of RDF and give a specification of this process in a functional language.
|
|
In the case
study the servers dispose of information files that contain facts and rules.
The facts are expressed in RDF. The syntax of the rules is RDF also, but the
semantic interpretation is different. Using rules implies inferencing. On top
of RDF a standardized inference layer has to be defined. A functional
language is very well suited for a declarative specification. At certain
points the inference process has to be interrupted to direct queries to other
sites on the World Wide Web. This implies a process of subquerying
where the inference process does not take place in one site (one computer)
but is distributed over several sites.
|
|
2) Is the
specified inference process sound and complete?
|
|
It is without
question of course that the answers to a query have to be valid. By
completeness is meant that all answers to a query are found. It is not always necessary that all
answers are found but often it is.
|
|
3) Which
kind of logic is best suited for the Semantic Web?
|
|
Using rules;
making queries; finding solutions; this is the subject of logic. So the
relevance of logic for an inferencing standard based on top of RDF has to be
investigated
|
|
4) What
can be done for augmenting the efficiency of the inference process?
|
|
A basic internet
engine should be fast because the amount of data can be high; it is possible
that data need to be collected from several sites over the internet. This
will already imply an accumulation of transmission times. A basic engine
should be optimized as much as possible.
|
|
5) How
can inconsistencies be handled by the inference process?
|
|
In a closed
world (reasoning on one computer system) inconsistencies can be mastered; on
the internet, when inferencing becomes a distributed process, inconsistencies
will be commonplace.
|
|
|