Using PSIs in inferencing PeterPaul Kruijsen Morpheus p

Using PSI’s in inferencing (Peter-Paul Kruijsen ~ Morpheus) p. kruijsen@mssm. nl Mail from pp to larsga (14 march 2005) "Is there a possibility to link a PSI to a predicate in an inference rule? " Imagine one topic map that defines a 'brother' association type with 'http: //psi. somedomain. org/brother' as PSI, and another topic map that defines only 'parenthood' associations with 'father', 'mother' and 'child' role players. In the second topic map an inference rule can be created: In the end, it could be about ontology mapping, i. e. mapping information between sources that use similar (= not equal) ontologies brother($X, $Y) : parenthood($MOTHER, $FATHER, $X), parenthood($MOTHER, $FATHER, $Y), $X /= $Y, male($X). • Another example (within same topic map) customer-provider(some-provider : provider, morpheus : customer) [p 1 : project = “Some project”] project-performance(p 1 : project, morpheus : performer) project-towards(p 1 : project, some-customer : customer) Could the two concepts of 'brother' be linked via PSI's, in other words, can a PSI be attached to a predicate in an inference rule. I could imagine creating only a 'brother' topic in the topic map with a PSI, and leaving it nontyping, but where does the logic kick in, saying the semantics are to be found in the inference rule. • Dynamic topic typing instance-of($P, composer) : instance-of($P, person), composed-by($P : composer). • Compare (various topic maps / ontologies) [john : person = "John"] {john, gender, "Male"} [john : person = "John"] gender-of(john, male) [john : male = "John"] super-sub(person : super, male : sub) More general question: "can inference rules be placed within a topic map? " Is there any logic in this? • Dynamic typing allows notion of interface from Java (Puccini of class person, implements composer) use those inference rules (via PSI) to constraints - opera should be composed by a composer - a project should be lead by an ontopian Mail from larsga to pp (14 march 2005) Very good question. I've been thinking about the relationship between inference rules and associations for years now without really finding a good way to unify them. I think you are right to want what you describe, but so far we don't have a good way to do this in tolog. I keep thinking about this and hoping to either get it into tolog or, at the very least, getting it right in TMQL. I think that's as much as I can tell you right now. • Requirements PSI's in inference rules tolog statements and inference rule heads with constants (PSI or strings) Use of constants in inference rule head Use of existing association-types in inference rule head Caching of inferred facts