Re: Slides "RuleML Meets RDF" for today's Joint Committee teleconand RDF's bNodes

From: pat hayes (phayes@ai.uwf.edu)
Date: 11/12/02

  • Next message: Mike Dean: "inference rules"
    >pat hayes wrote:
    >...
    >>  >Looking at the RDF Model Theory edited by Pat, we are still open to various
    >>  >possibilities for "simply indicating the existence of a thing, without
    >>  >using,
    >>  >or saying anything about, the name of that thing"
    >>  >(http://www.w3.org/TR/rdf-mt/#unlabel).
    >>  >
    >>  >It now is clear to me that anonymous logical variables are no solution,
    >>  >since
    >>  >they are universally, not existentially, interpreted in facts.
    >>
    >>  Doesnt that depend on the logic you are using?
    >
    >I was talking about rules and facts in Horn logic.
    >To give an example in pure Prolog syntax, instead of the ground fact
    >
    >has-creator("http://www.w3.org/Home/Lassila","Ora Lassila").
    >
    >someone could assert the non-ground fact
    >(using "_" as the anonymous logical variable)
    >
    >has-creator(_,"Ora Lassila").
    >
    >This is equivalent to the non-ground fact
    >(using "X" as a named logical variable that is fresh in this clause)
    >
    >has-creator(X,"Ora Lassila").
    >
    >X is interpreted universally, not existentially.
    
    As I say, that interpretation depends on the logical rules one is 
    using. However, it is usual to interpret free variables as universals 
    in a 'rule' language, I agree.
    
    >So, anonymous logical variables in Prolog facts are unrelated
    >to anonymous nodes (bNodes) in RDF triples.
    >
    >In Prolog queries, however, anonymous logical variables are interpreted
    >existentially.
    
    Right, the query always inverts. Gentzen was a clever old fellow.
    
    >
    >>  >I now think RDF's bNodes are more akin to some of the many uses of null
    >>  >values in
    >>  >relational databases or, perhaps, closer to RDF, in object-relational
    >>  >databases.
    >>  >A quick Google search revealed a related remark in "Topic Maps and RDF"
    >>  >(http://www.isotopicmaps.org/pipermail/sc34wg3/2002-August/000495.html).
    >>  >
    >>  >Null values are absent from Prolog, but in Dagstuhl discussions with Jens
    >>  >Dietrich
    >>  >(http://www.jbdietrich.de/about.html) we thought they would be important in
    >>  >RuleML
    >>  >anyway (he is working on RuleML extensions for relational databases).
    >>
    >>  Harold, please, calm yourself. They are just existential variables.
    >>  Logic has had them since about 1878.
    >
    >Yes, I started with this at the beginning and came back to this at the end:
    >It's the most logical interpretation of RDF's bNodes. But I think 
    >the excursion
    >into relational databases is a worthwhile one, if only because some RDF triple
    >stores have been interfaced to, or implemented in, relational databases.
    >
    >I guess that you could generalize your RDF analysis to relational databases
    >and (re?)discover that some of the many uses of null values are just 
    >existential
    >variables.
    >
    >>  >Yet, this would not give us the local existential scope of RDF's bNodes.
    >>  >We hope our analysis of the RDF Model Theory will lead to a good solution
    >>  >for
    >>  >RuleML bNodes as well.
    >>  >
    >>  >Perhaps we should also have a closer look at eigenvariables in 
    >>Lambda-Prolog
    >>  >(http://www.cse.psu.edu/~dale/lProlog/).
    >>  >Maybe we can even learn something here from the discussion of the "line of
    >>  >identity"
    >>  >in Charles Sanders Peirce's Existential Graphs
    >>  >(http://users.bestweb.net/~sowa/peirce/ms514.htm).
    >>
    >>  Lets not get crazy here. There isn't anything very deep or mysterious
    >>  about existential quantifiers.
    >
    >No, but perhaps in their diverse graphical representations. Since RDF's graph
    >models are so important to the RDF and Semantic Web community, 
    >perhaps it would
    >be worthwhile to still try to optimize these graphs a bit.
    
    Sigh. The chief argument often made for the RDF graph syntax is that 
    it is optimal for software to use.
    
    >
    >>  Pat
    >
    >>  >Ultimately, we may need explicit (existential) quantifiers for glueing
    >  > >together
    >>  >a conjunction (or a rulebase) of facts.
    >
    >To be specific, we may need rulebases like these:
    >
    >... normal facts and rules ...
    >(Ex X
    >   has-creator(X,"Ora Lassila").
    >   was-accessed-by(X,"Ora Lassila").
    >)
    >... normal facts and rules ...
    
    Right.
    
    I am a bit at a loss to know what the fuss is about. The actual logic 
    involved here seems completely obvious. RDF graphs are 
    existential-conjunctive assertions. 'Rules" is a rather misleading 
    term for a universal implication. All the required logic follows from 
    this.
    
    Isnt there something more useful to discuss?
    
    Pat
    -- 
    ---------------------------------------------------------------------
    IHMC					(850)434 8903   home
    40 South Alcaniz St.			(850)202 4416   office
    Pensacola              			(850)202 4440   fax
    FL 32501           				(850)291 0667    cell
    phayes@ai.uwf.edu	          http://www.coginst.uwf.edu/~phayes
    s.pam@ai.uwf.edu   for spam
    


    This archive was generated by hypermail 2.1.4 : 11/12/02 EST