Menu

#246 Errors in KRSS2 renderer and parser

open
None
5
2012-10-13
2011-02-11
No

The output of the KRSS2 renderer is erroneous and cannot be read by the KRSS2 parser. These are the issues (R for "renderer", P for "parser"):

(R1) Disjointness axioms seem to be completely omitted, for classes and properties.
(R2) Equivalence axioms for properties are omitted.
(R3) Inverse statements (InverseObjectProperties(r, rInv)) are omitted.
(R4) General class axioms (GCIs) with equivalences are omitted.
(R5) Some general class axioms are rendered without "implies".
(R6) When range restrictions of an object property are rendered, they contain the classes from its domain restrictions instead.
(R7) Not a bug, but difficult to handle for external parsers: GCIs and axioms about object properties are rendered in one large line.

(RP1) The renderer writes "::parents", "::domain", "::range", but the parser expects ":parents", ":domain", ":range".

(P1) Occurrences of "inverse" cause errors.
(P2) ":parent" constructs following ":range" or ":domain" constructs cause errors.
(P3) Not a bug, but a redundancy: after parsing, the ontology contains axioms "A <= owl:Thing" for every A -- and even "owl:Thing <= owl:Thing". This gives no information and can be avoided by creating declaration axioms instead of subclass axioms.

I understand that some repairs to the renderer will require the use of constructs that are not in the standard KRSS specification (R1 for properties, R2, R3). For others, a workaround can be found in the standard vocabulary, e.g., writing equivalences as biimplications. I think that the KRSS2 renderer/parser should introduce new vocabulary for these cases and otherwise use the standard vocabulary.

A small example that contains the above cases except R5 is attached as krsstest2 twice: the original ontology in RDF/XML and the output of the krss renderer. I can provide a larger example that exhibits R5 on request -- I don't have permission to make the ontology public.

I'm happy to discuss these issues and/or help fixing them, but I don't want to become active without Olaf's consent.

Discussion

  • Thomas Schneider

    archive containing small example in .owl and .krss

     
  • Ignazio Palmisano

    • assigned_to: nobody --> ignazio1977
     

Log in to post a comment.

MongoDB Logo MongoDB