Re: Cross-ontologies reasoning

To answer Bijan's question:

On Dec 11, 2003, at 2:43 PM, Bijan Parsia wrote:

> On Thursday, December 11, 2003, at 05:00 PM, Francis McCabe wrote:
>
>> Ehem
>>   Notwithstanding the technologies being discussed, *translation* 
>> between ontologies is about as tractable in the general case as 
>> mapping between English and Japanese.
>
>>  However, an approach based on translating queries is, in principle, 
>> more doable than arbitrary mapping.
>
> Care to articulate the difference that makes the difference in your 
> eyes?


I do not have the energy to prove this, but there is an analogy with 
the halting problem. The corollary to the halting problem is, of 
course, that any *given* program/data set pair can be checked to see if 
it halts (by simply running it).

In the case of Ontologies, I think that in practice it makes a 
significant difference to map a single query from one Ontology to 
another; compared to translating the entire ontology. The reasoning is 
that in mapping a single query you are projecting a one-dimensional 
vector from the foreign ontology to the target ontology. In simpler 
terms, you only need to verify the relatively small fragment uncovered 
by a query is correctly interpretable in the target ontology.

Of course, I am aware that it is possible to construct a query like 
"tell me all you know" which will reduce to the general case. But, 
again, in practice, this is unlikely. A far more likely query is "what 
does this customer's address" mean in your system? That way, if there 
are concepts in the foreign system that are unknown to the target, then 
they will not normally 'show up' in the queries originating from the 
target system.
Frank

Received on Wednesday, 17 December 2003 00:06:44 UTC