Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: Brainstorming Idea: Set up an online testbed



Hi, John,

A few other questions inspired, in part, by your questions - hopefully 
relevant to the problem of stating requirements for systems that support 
interoperability.

1) Just what do we *mean* by 'interoperability'?  The term itself has 
been kicked around in a number of venues for close on a decade, but I am 
not familiar with any attempt to formally specify what it means for two 
systems (ontologies, inference schemes) to interoperate.  

2) Assuming a formal specification of interoperability that supports a 
metric of how well two systems/ontologies/inference schemes interoperate 
with each other, are there any limitations on what is provable regarding 
prospective interoperability under given conditions of use or for given 
problem domains?

3) When mapping one ontology into another, can we characterize the kinds 
of adjustment that may be necessary, and can we say anything about the 
computational costs associated with making these adjustments?

4) Especially as regards your question '2':  human cultures have also 
been self-organizing for millenia, and humans have been co-evolving for 
quite a bit longer than that, within the framework of a biosphere within 
which evolutionary histories are determinative of salience.  What 
scaffolding, if any, does a process such as this afford for the 
alignment of ontologies and mental models, what conditions does it 
impose, and is it in any way important to the process and/or options for 
alignment?

Ben Rode


John F. Sowa wrote:
> I strongly agree with Mehrnoush Shamsfard:
>
> > I still believe that working to create a technology to map
> > various ontologies with various perspectives is more important
> > than creating standards. Maybe the core principles or beliefs
> > or concepts or ... should be the same in various ontologies
> > (I'm not sure about) I just know that none of us has a unique
> > standard mental ontology same as the others, but we can more
> > or less understand each other. It seems that what we have in
> > common is not the standard upper ontology as we use in computer
> > systems. Discussing this matter can help to create ontologies
> > which are not standard but can be used for interoperability.
>
> The requirement that two systems must have *identical* ontologies
> before they can work together is unrealistic.
>
> For example, we can all interoperate by email to this list,
> despite the fact that we all have very different background
> knowledge and points of view.  And computers have been working
> together across various networks for over 40 years without
> having explicit ontologies.
>
> Before we can state requirements for ontologies that support
> interoperability, we must determine what enables current
> systems to interoperate and what more would be useful.
>
> Some questions:
>
>  1. How do current systems, even without formal ontologies,
>     interoperate?  What is required for those operations?
>     Is there a single kind of requirement or many different
>     kinds for different kinds of systems or applications?
>
>  2. People have been working together for millennia without
>     aligning everything they know before they even start.
>     How do they work together?  What can we learn from that?
>
>  3. What improvements would a common ontology confer?
>     For what kinds of systems or applications?
>
>  4. There are many ontologies being developed today with little
>     or no coordination among them.  How can systems based on
>     them interoperate with each other and with legacy systems?
>
> These are some of the questions that must be addressed before
> we can formulate any realistic requirements for ontologies.
> Without some idea of what we mean by interoperability, it is
> pointless to try to establish standards to support it.
>
> John Sowa

begin:vcard
fn:Dr. Benjamin Rode
n:Rode;Dr. Benjamin
org:Cycorp;Senior Technical Staff
adr:Suite 250;;7718 Wood Hollow Drive;Austin;TX;78731-1601;USA
email;internet:ben@cyc.com
title:Knowledge Representation Specialist
tel;work:512-342-4022
tel;fax:512-340-4040
url:http://www.cyc.com
version:2.1
end:vcard