On 4/14/07, Stephen Reed <[EMAIL PROTECTED]> wrote:
While at Cycorp I was responsible for two iterations of producing OWL from
the Cyc knowledge base. The second iteration also used the SWRL proposal which builds upon OWL to represent horn clause rules. Here are features/issues that might interest you.
1. OWL represents ground binary propositions only - no explicit negation
operator, no explicit for-all or there-exists quantification, no variables.
2. OWL does have class and property multiple inheritance. 3. OWL terms are atomic and do not represent FOPL function terms. 4. OWL does not have the unique names assumption. For each unique term
in your KB you should include it in a list of all-different terms.
5. OWL does not have a notion of context. 6. The creators of OWL were familiar with the Cyc Public (upper)
Ontology. Furthermore OWL is based on RDF which was developed by R. Guha who is a former Cyc Project member. So OWL is very compatible with the Cyc ontology with respect to class (Cyc Collection) and property (Cyc predicate) subsumption.
7. Multiple inheritance is permitted, like Cyc. 8. The class hierarchy is singly rooted in Thing, like Cyc. 9. Classes themselves can be members of other classes, like Cyc. 10. Disjointness in OWL works exactly like disjointness (non-intersecting
Collections) in Cyc.
As perhaps you know my own work is based on the OpenCyc ontology, merged
with WordNet, the CMU Pronouncing Dictionary and Wikitionary. Looking forward to OWL compatibility I have designed my system to only create atomic terms and binary propositions. I do have a numeric strength on each proposition that is not directly expressible in OWL.
Is this helpful? -Steve
Thanks - very helpful =) Maybe we can collaborate on some project since our approaches are similar. I'm interested in collecting facts and rules via the internet to form a KB, possibly larger than Cyc's, but I'm also considering collaborating with Cycorp. Why did you decide to develop something on your own, the Texai? YKY ----- This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415&user_secret=fabd7936
