log:conjunction and named graphs

[Henry Story-2]

I don't seem to be able to get the tests for log:conjunction to work  
when altered to use the names of graphs.

------------
$ cat test2.n3
@prefix log: <http://www.w3.org/2000/10/swap/log#>.
@prefix owl: <http://www.w3.org/2002/07/owl#>.
@prefix : <#> .

# This gives the correct conclusions
#{ ( { :sky :color :blue }  { :grass :color :green } )  
log:conjunction ?F }
#  =>  { :rest :is ?F } .


#but this does not
:g = { :sky :color :blue } .
:h = { :grass :color :green } .


{ ( :g :h ) log:conjunction ?F } => { :rest :is ?F } .

------------

If I then run

cwm test2.n3 --think --apply=/Users/hjs/Programming/w3.org/swap/test/
owl/owl-rules.n3 --think

I don't get the

rest is { :sky color :blue . :grass :color :green . }

result I was expecting.

Any ideas?

Henry

[Tim Berners-Lee]

Henry,

You are using owl:sameAs.

cwm does not do owl:sameAs processing by default.

It is an RDF processor, so it must be able to handle and input and output owl:sameAs graphs without processing

There is some optional owl:sameAs processing with --closure=e  which does some smushing -- not much.

This will not, however, smush a names node onto a literal including a literal graph.

It doesn't do = inference in backward chaining in the antecedent of a query.

Tim

$ cwm --help

Command line RDF/N3 tool

        

 <command> <options> <steps> [--with <more args> ]

options:

 

--pipe        Don't store, just pipe out *

steps, in order left to right:

--rdf         Input & Output ** in RDF/XML insead of n3 from now on

--n3          Input & Output in N3 from now on. (Default)

--rdf=flags   Input & Output ** in RDF and set given RDF flags

--n3=flags    Input & Output in N3 and set N3 flags

--ntriples    Input & Output in NTriples (equiv --n3=usbpartane -bySubject -quiet)

--language=x  Input & Output in "x" (rdf, n3, etc)  --rdf same as: --language=rdf

--languageOptions=y     --n3=sp same as:  --language=n3 --languageOptions=sp

--ugly        Store input and regurgitate, data only, fastest *

--bySubject   Store input and regurgitate in subject order *

--no          No output *

              (default is to store and pretty print with anonymous nodes) *

--base=<uri>  Set the base URI. Input or output is done as though theis were the document URI.

--closure=flags  Control automatic lookup of identifiers (see below)

<uri>         Load document. URI may be relative to current directory.

--apply=foo   Read rules from foo, apply to store, adding conclusions to store

--patch=foo   Read patches from foo, applying insertions and deletions to store

--filter=foo  Read rules from foo, apply to store, REPLACING store with conclusions

--query=foo   Read a N3QL query from foo, apply it to the store, and replace the store with its conclusions

--sparql=foo   Read a SPARQL query from foo, apply it to the store, and replace the store with its conclusions

--rules       Apply rules in store to store, adding conclusions to store

--think       as -rules but continue until no more rule matches (or forever!)

--engine=otter use otter (in your $PATH) instead of llyn for linking, etc

--why         Replace the store with an explanation of its contents

--why=u       proof tries to be shorter

--mode=flags  Set modus operandi for inference (see below)

--reify       Replace the statements in the store with statements describing them.

--dereify     Undo the effects of --reify

--flatten     Reify only nested subexpressions (not top level) so that no {} remain.

--unflatten   Undo the effects of --flatten

--think=foo   as -apply=foo but continue until no more rule matches (or forever!)

--purge       Remove from store any triple involving anything in class log:Chaff

--data              Remove all except plain RDF triples (formulae, forAll, etc)

--strings     Dump :s to stdout ordered by :k whereever { :k log:outputString :s }

--crypto      Enable processing of crypto builtin functions. Requires python crypto.

--help        print this message

--revision    print CVS revision numbers of major modules

--chatty=50   Verbose debugging output of questionable use, range 0-99

--sparqlServer instead of outputting, start a SPARQL server on port 8000 of the store

--sparqlResults        After sparql query, print in sparqlResults format instead of rdf

finally:

--with        Pass any further arguments to the N3 store as os:argv values

 

            * mutually exclusive

            ** doesn't work for complex cases :-/

Examples:

  cwm --rdf foo.rdf --n3 --pipe     Convert from rdf/xml to rdf/n3

  cwm foo.n3 bar.n3 --think         Combine data and find all deductions

  cwm foo.n3 --flat --n3=spart

Mode flags affect inference extedning to the web:

 r   Needed to enable any remote stuff.

 a   When reading schema, also load rules pointed to by schema (requires r, s)

 E   Errors loading schemas of definitive documents are ignored

 m   Schemas and definitive documents laoded are merged into the meta knowledge

     (otherwise they are consulted independently)

 s   Read the schema for any predicate in a query.

 u   Generate unique ids using a run-specific

Closure flags are set to cause the working formula to be automatically exapnded to

the closure under the operation of looking up:

 s   the subject of a statement added

 p   the predicate of a statement added

 o   the object of a statement added

 t   the object of an rdf:type statement added

 i   any owl:imports documents

 r   any doc:rules documents

 E   errors are ignored --- This is independant of --mode=E

 n   Normalize IRIs to URIs

 e   Smush together any nodes which are = (owl:sameAs)

See http://www.w3.org/2000/10/swap/doc/cwm  for more documentation.

Setting the environment variable CWM_RDFLIB to 1 maked Cwm use rdflib to parse

rdf/xml files. Note that this requires rdflib.

Flags for N3 output are as follows:-

        

a   Anonymous nodes should be output using the _: convention (p flag or not).

d   Don't use default namespace (empty prefix)

e   escape literals --- use \u notation

i   Use identifiers from store - don't regen on output

l   List syntax suppression. Don't use (..)

n   No numeric syntax - use strings typed with ^^ syntax

p   Prefix suppression - don't use them, always URIs in <> instead of qnames.

q   Quiet - don't output comments about version and base URI used.

r   Relative URI suppression. Always use absolute URIs.

s   Subject must be explicit for every statement. Don't use ";" shorthand.

t   "this" and "()" special syntax should be suppresed.

u   Use \u for unicode escaping in URIs instead of utf-8 %XX

v   Use  "this log:forAll" for @forAll, and "this log:forAll" for "@forSome".

/   If namespace has no # in it, assume it ends at the last slash if outputting.

Flags for N3 input:

B   Turn any blank node into a existentially qualified explicitly named node.

Flags to control RDF/XML output (after --rdf=) areas follows:

        

b  - Don't use nodeIDs for Bnodes

c  - Don't use elements as class names

d  - Default namespace supressed.

l  - Don't use RDF collection syntax for lists

r  - Relative URI suppression. Always use absolute URIs.

z  - Allow relative URIs for namespaces

    Flags to control RDF/XML INPUT (after --rdf=) follow:

        S  - Strict spec. Unknown parse type treated as Literal instead of error.

        T  - take foreign XML as transparent and parse any RDF in it

             (default it is to ignore unless rdf:RDF at top level)

        L  - If non-rdf attributes have no namespace prefix, assume in local <#> namespace

        D  - Assume default namespace decalred as local document is assume xmlns=""

        R  -  Do not require an outer <rdf:RDF>, treating the file as RDF content (opposite of T)

    Note: The parser (sax2rdf) does not support reification, bagIds, or parseType=Literal.

          It does support the rest of RDF inc. datatypes, xml:lang, and nodeIds.

$ 

On 2009-04 -17, at 09:52, Henry Story wrote:

I don't seem to be able to get the tests for log:conjunction to work when altered to use the names of graphs.

------------
$ cat test2.n3
@prefix log: <http://www.w3.org/2000/10/swap/log#>.
@prefix owl: <http://www.w3.org/2002/07/owl#>.
@prefix : <#> .

# This gives the correct conclusions
#{ ( { :sky :color :blue }  { :grass :color :green } ) log:conjunction ?F }
#  =>  { :rest :is ?F } .


#but this does not
:g = { :sky :color :blue } .
:h = { :grass :color :green } .


{ ( :g :h ) log:conjunction ?F } => { :rest :is ?F } .

------------

If I then run

cwm test2.n3 --think --apply=/Users/hjs/Programming/w3.org/swap/test/owl/owl-rules.n3 --think

I don't get the

rest is { :sky color :blue . :grass :color :green . }

result I was expecting.

Any ideas?

Henry

[Dan Connolly]

On Fri, 2009-04-17 at 13:39 -0400, Tim Berners-Lee wrote:

> Henry,
>
>
> You are using owl:sameAs.
>
>
> cwm does not do owl:sameAs processing by default.
>
>
> It is an RDF processor, so it must be able to handle and input and
> output owl:sameAs graphs without processing
>
>
> There is some optional owl:sameAs processing with --closure=e  which
> does some smushing -- not much.

you might also try mixing in some rules about owl:sameAs:

{ ?u = ?uu. ?u ?p ?v. } => { ?uu ?p ?v }.
{ ?u ?p ?v. ?v = ?vv } => { ?u ?p ?vv }.

That's taken from
  http://www.w3.org/2000/10/swap/util/owlth.n3

it's a cwm-happy elaboration of rdfp11
from ter-Horst's work.

# H.J. ter Horst, Completeness, Decidability and Complexity of
Entailment for RDF
# Schema and a Semantic Extension Involving the OWL Vocabulary, Revised
and
# extended version of [11], Journal of Web Semantics 3 (2005) 79-115.
# http://www.websemanticsjournal.org/ps/pub/2005-15


--
Dan Connolly, W3C http://www.w3.org/People/Connolly/
gpg D3C2 887B 0F92 6005 C541  0875 0F91 96DE 6E52 C29E

[Dan Connolly]

On Sun, 2009-04-19 at 17:51 -0500, Dan Connolly wrote:
> On Fri, 2009-04-17 at 13:39 -0400, Tim Berners-Lee wrote:

Never mind what I said:

> you might also try mixing in some rules about owl:sameAs:

I tested it and it doesn't help.

What Tim said is more relevant:

> > There is some optional owl:sameAs processing with --closure=e  which
> > does some smushing -- not much.

You might try one of the backward chainers, i.e. Euler...
maybe fuxi would help. But as you're quantifying over
formulas, those engines might not help either.

--
Dan Connolly, W3C http://www.w3.org/People/Connolly/
gpg D3C2 887B 0F92 6005 C541  0875 0F91 96DE 6E52 C29E