IW Meeting 2007 Feb 28

From Inference Web

Jump to: navigation, search

agenda

  • IW2
    • general: new cvs directory (iw2), Sourceforge (inferenceweb); persistent URI
    • PML2API progress:
      • handle subclass relation:
        • jastor: interface extends super-interface1 (, interface2); class extends JASTOR.Thing implements interface
        • us: interface extends super-interface; class extends super-class implements interface
        • [li] our approach may encounter conflicting method implementation by inheriting methods from super-class. e.g. superclass has the method (card<=1), but subclass does not (card=0)
        • [alyssa] we are using object oriented approach
      • handle ontologies in old version
        • A: during loading property-value pair
        • B: translate to v2 first (in RDF model) and then process it in v2 way
        • [li] B is better if complete translation is available.
        • [alyssa] does not matter to end developers
      • handle methods of java class
        • jastor/protege/PML1: getPropertyABC(), setPropertyABC(value), addPropertyABC(), ... each property will cause a set of methods, good for java IDE developers.
        • us: getProeprty(ABC), setProperty(ABC, value), one method handle everything, require users know PML vocabulary in advance.
        • [li] jastor approach seems more handy while ours is more flexible
        • [alyssa] prefer the latter, assume api users are kind of knowledgeable with PML ontology
      • offline mode -- requested by GILA
      • junit test
    • PML2 ontology progress: last call for review
      • pml-owl.owl: moved hasResourceReference, Do we need a new subclass of Information? added more specific domain specification
      • need documentation
  • GILA
    • ontology release for iteration 2
    • need to deliver new PML2 ontology and API
    • started discussion on preference ontology
  • TAMI
    • scenario 9 revised
    • helping CPH
    • cwm-pml translation
  • WineAgent
    • mayukh encounter "default URI" issue.

action items

  • Deborah will be absence for the next meeting, Li will host the meeting
  • PML API: performance test. GILA's "PML delay", JTP may cause delay, Jena OntModel delay.
  • GILA simulation
    • RA1 installed
    • lisp installtion
    • run the simulation


PML related issues

  1. (PML ontology) decision on declarativeRule/methodRule, see deborah's email
  2. (PML API) translation from PML ontology to java
    • there are three assumptions involved in the translation
      1. [single most specific type] given an instance I, which is an instance of owl:Thing, we can always find its most specific class, such that if there exist classes C1 and C2 (I rdf:type C1)(I rdf:type C2) then it must be true either C1 is subclass of C2 or C2 is subclass of C1
        • this feature enables us to find the most specific java class of any owl instance
        • [suggestion] we need this assumption
      2. [no multiple inheritance] given a class C0, all its super classes are connected via subclass relation, such that if there exist classes C1 and C2 (C0 subclass C1) (C0 subclass C2) then it must be true either C1 is subclass of C2 or C2 is subclass of C1
        • this assumption is imposed when we translate subclass relation to java class-inheritence hierarchy.
        • However, it can be avoided by using interface inheritence without class-inheritence
        • [suggestion] we need to remove this assumption
      3. [domain of property must be defined in the same class] given an ontology ont1 defining a property p1, then the domain c1 of p1 must also be defined in ont1.
        • this assumption is imposed when we translate an ontology to a java package. If c1 is defined in another ontology ont2, the orginal version (java class) of C1 must stay in the java package corresponding to ont2, while the extended version (java class) of c1 should go to java package corresponding to ont1.
        • [suggestion] this assumption should be removed because we need both the orginal version and extended version.
  • (PML API) data manipulation
    • no convertion during reading PML data,
    • create data with PML2,
    • dump by default using the current verison, otherwise by specified version
  • hassel in accessing data encoded in different versions of ontology
    • there are two alternative approaches:
      • (i) jastor/protege/PML1: getPropertyABC(), setPropertyABC(value), addPropertyABC(), ... each property will cause a set of methods, good for java IDE developers.
      • (ii) getProeprty(ABC), setProperty(ABC, value), one method handle everything, require users know PML vocabulary in advance.
    • it seems that we need both, the first one requires understanding of PML, and possibly older versions if such data will be encountered, therefore, users of PML API may need to decide using get("hasAntecedentList") or get("hasAntecedent") to when encountering an instance of IWInferencestep according to its namespace. The latter will hide this decision, but also need intenstive programming work.
Personal tools
Navigation