ISSUE-5: n-ary datatypes - decidability on merging

4) Jeff Pan's design does not ensure that decidability is preserved on
merging

5) Problems with decidability on merging are *very* likely, (or require
new research to show how to avoid them)

===

One of the motivating examples for n-ary datatypes is unit conversion.

Unit conversion is (only?) needed when merging data from more than one 
source.

In Jeff Pan's design for addressing the decidability issues (which is 
the only design I am aware of), he shows that when merging data using 
two different ontologies you need to merge the datatype groups.

In order for the reasoning task to be decidable the datatype group needs 
to be 'conforming'.

Unfortunately, the merge of two conforming datatype groups is not 
necessarily conforming. In fact, there are obvious and realistic cases 
in which it is not.

Pan and Horrocks do show that the merge of two non-intersecting 
conforming datatype groups is conforming, but this result is practically 
useless, since every datatype group of interest includes string, and 
most include integer.

Problems are very likely,
since a datatype group including:
- multiplication
- addition
- integer
- inequality

is not conforming but many of the subgroups are. These four datatypes 
are sufficiently basic to be widely occurring, and, without guidance as 
to the systematic avoidance of some of them, users are likely to use these.




Jer

Received on Monday, 12 November 2007 20:16:01 UTC