Re: [ontolog-forum] {quality-methodology} Architectural considerations i

[Amanda Vizedom <amanda.vizedom@xxxxxxxxx>]

I am copying this response to a thread on [ontolog-forum] also to [ontology-summit], as it relates to similarly titled thread there, and to the question of what ontology features are important for what kinds of of ontology applications.

Three points below: (1) on the integrating vs. reasoning view; (2) a  correction regarding statements that Cyc ontology is not meant for  integrating; and (3) regarding what appears to be a talking-past, regarding context (in particular, regarding the possibility of representing and reasoning across different contexts.

(1) Matthew West said [MW]:

Dear John,

Unfortunately, both you and Doug completely missed the point that we were

talking about methodologies for INTEGRATING (or conceptual) ontologies, not

reasoning ontologies.

I'd like to reel that line of argument in a bit. It can be useful to distinguish between integration-centric (or interoperability-centric) uses of ontology and reasoning-centric uses of ontology.

However, to assume that such uses must, or usually do, take place in separate applications, or be performed using different ontologies, is to step beyond the information at hand. And to step right into the knowledge gap around which kinds of ontologies are used successfully for what.

In fact, much ontology-based information integration or federation requires reasoning to some degree or other. Sometimes the reasoning is quite limited, e.g., to class subsumption and similar. Sometimes the reasoning is quite sophisticated, as in applications using Cyc's Semantic Knowledge Source Integration capabilities (about which, more in a moment).  Often it's somewhere in the middle: Generally subsumption based, with areas of greater depth and sophistication.  The point is that it's quite common that ontologies intended to support information interoperability must support some reasoning in order to perform their primary goal. 

(2) Regarding the various comments that have been made about Cyc in this context, a few corrections are needed. First of all, to the claim that Cyc is not an integrating ontology. This is simply false. Cyc is often used to do integration and reasoning in the same application. In fact, this combining capability existed and was used in Cyc projects almost 20 years ago, and was called out, named, and marketed as a high-value use of Cyc in the early 2000s.  [See, for example: Masters J, Güngördü Z.  2003.  Structured Knowledge Source Integration: A Progress Report . Integration of Knowledge Intensive Multiagent Systems, Cambridge; Masters J.  2002.  Structured Knowledge Source Integration and its applications to information fusion . Fifth International Conference on Information Fusion. These are downloadable from http://www.cyc.com/publications. Not on that page, but also relevant, is a paper I presented at Fusion 2001: "Knowledge Base Support for Decision Making Using Fusion Techniques in a C2 Environment." This paper described work for DARPA's Command Post of the Future project, using integration and reasoning via Cyc to support detection of anomaly patterns suggestive of enemy intent, where such detection required processing input from multiple (HUMINT and SIGINT) information streams and reasoning over the combined information.]  The point here is that not only is integration within Cyc's intended purposes, but such use was well out in front of the curve of semantic interoperabilty.

(3) Doug and John made comments about use of Cyc microtheories in this regard.  Matthew objects with several ways of putting the point:

MW: It is the other way round. We need to relate the concepts in a wide range of systems to the normative standards. That gives some interesting requirements on those standards:
 - The standard needs to be able to have any concept from those legacy system mapped to it, ...

 

and

MW: On the contrary, for an integrating ontology, you only want one way to say something, but you want it to be the most expressive way of saying it, so that any other way can be expressed that way.

and

 

MW: That is even worse. You have to be able to translate between conflicting representations, and bring the results into one environment, or you have not actually managed to integrate anything.

 

MW: No. What you would need is a mapping between Cyc and your integrating ontology so that different microtheories that covered the same area of the integrating ontology were mapped to the same place in the integrating ontology. This would bring all the ground facts in Cyc into a form where they could be compared, and integrated with those from different sources.

I think there is a talking-past here. This might be because of some undescribed features of Cyc's microtheories. In particular, they are not dumb containers; they can be related to one another in all sorts of ways, and reasoning across them is common. You can even specify: that a particular source external to Cyc (agent, db, or other) is the expert about particular areas of knowledge; that a particular microtheory matches or models that source's ontology; that particular assumptions (however upper, mid, or lower level) hold within this microtheory, that particular relationships hold between the things in that microtheory and particular things in other (typically, more general) microtheories, and so on.  This enables Cyc to exchange information with the source within its own model, and then to relate the resulting information to other information in Cyc, whether it be stored in Cyc or accessed from another such source 

If I'm understanding  Matthew correctly, this gives the desired integrating functionality. That is, given how microtheories work in Cyc, it doesn't appear that there is much actual, or deep, disagreement here.

Best,

Amanda

P.S. As usual, since it has been almost a decade since I worked at Cycorp, I encourage current (or more recent) #$Cyclists to correct or update what I said, and to provide more or better references and examples.

_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/  
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/ 
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J    (01)