W3C home > Mailing lists > Public > public-prov-comments@w3.org > January 2013

Re: PROV comments from Clark&Parsia

From: Evren Sirin <evren@clarkparsia.com>
Date: Wed, 23 Jan 2013 11:24:02 -0500
Message-ID: <CAFC3-QpnWSOuEpQaV+1zzqpWBb0YWUPgCY102qGFoycFjhkpjA@mail.gmail.com>
To: Paul Groth <p.t.groth@vu.nl>
Cc: "public-prov-comments@w3.org" <public-prov-comments@w3.org>, Héctor Pérez-Urbina <hector@clarkparsia.com>
Hi Paul,

The short summary is we are fine with all the responses from the WG
except the response for "ISSUE-611 (comments on prov-o)" for which we
request more clarification. See below for more details.

On Mon, Jan 21, 2013 at 4:56 AM, Paul Groth <p.t.groth@vu.nl> wrote:
> Dear Evren,
>
> First, thanks for your interest in PROV. We are really excited that Clark &
> Parsia are looking at implementing PROV within Stardog.
>
> The working group has looked at the issues you raised and produced replies.
> We divided your comment into four parts. You'll find the responses linked to
> below.
>
> - ISSUE-611 (comments on prov-constraints)

These changes address our concerns.

>
> - ISSUE-611 (PROV-CONSTRAINT Test Cases)

We are happy with the changes to the tests and the explanation.

>
> - ISSUE-611 (comments on prov-o)

Our comment was not regarding encoding of the constraints in OWL
(which is not possible to do completely anyway) but about encoding the
inferences in OWL. Right now, it looks like some of the inferences
from PROV Constraints document is included in PROV-O. Specifically,
Inference 15 (influence-inference) [1] and Inference 20
(specialization-alternate-inference) [2] are included in PROV-O as
subPropertyOf axioms. But other inferences defined in this document
are not included in PROV-O which is a little confusing. For example,
Inference 12 (revision-is-alternate-inference) [3] suggests another
subPropertyOf relation (wasRevisionOf subPropertyOf alternateOf) but
this is not in PROV-O. If the WG chooses to encode some of the
inferences in PROV-O but not others, we would like to understand the
rationale behind this decision.

>
> - ISSUE-611 (Test cases for other specifications)

Our comment was regarding test cases that test different aspects of
PROV specification other than validation such as inferences. Analogous
to OWL test cases where there are tests of (in)consistency and
(non-)entailment, there could be PROV test cases for just testing
entailments (transitivity, inverses, etc. as defined in the PROV
constraints document) in addition to the current tests that check
(in)validity. Having said that, this comment is meant to be a
recommendation for additional future tests and we are fine with the
current set of tests.

>
> We hope these address your concerns. If you could do us a favour and
> acknowledge that these responses either address concerns or do not.

Best,
Evren

[1] http://www.w3.org/TR/prov-constraints/#influence-inference
[2] http://www.w3.org/TR/prov-constraints/#specialization-alternate-inference
[3] http://www.w3.org/TR/prov-constraints/#revision-is-alternate-inference

>
> Thanks again for your detailed comments,
> Paul
>
>
>
>
>
>
>
>
> On Fri, Jan 4, 2013 at 6:23 PM, Evren Sirin <evren@clarkparsia.com> wrote:
>>
>> Hi all,
>>
>> We are working towards supporting PROV inferences and constraints in
>> our RDF database Stardog [1]. Below are some comments about PROV
>> specification documents that we identified while working on our
>> implementation.
>>
>> * PROV Constraints
>>
>> 1. The flow control arrows in Figure 1 seem to be backwards.
>> 2. Definition 2.1 seems to be missing the id on the right-hand side.
>> 3. Since uniqueness constraints are ‘applied’ and can derive new
>> atoms, it is misleading to call them constraints. The same applies to
>> typing constraints.
>> 4. The definition of enforcement of uniqueness constraints states one
>> should apply the resulting substitution to the whole PROV instance.
>> However, the scope of the variables is not sets of rules.
>> 5. Inference 9 (wasStartedBy-inference) should be: IF
>> wasStartedBy(_id; _a,e1,a1,_t,_attrs), THEN there exist _gen and _t1
>> such that wasGeneratedBy(_gen; e1,a1,_t1,[]).
>> 6. Inference 10 (wasEndedBy-inference) should be: IF wasEndedBy(_id;
>> _a,e1,a1,_t,_attrs), THEN there exist _gen and _t1 such that
>> wasGeneratedBy(_gen; e1,a1,_t1,[]).
>> 7. Inference 15.4 should be: IF wasStartedBy(id; a2,e,_a1,_t,attrs)
>> THEN wasInfluencedBy(id; a2, e, attrs).
>> 8. Inference 15.7 should be: IF wasDerivedFrom(id; e2, e1, _a, _g, _u,
>> attrs) THEN wasInfluencedBy(id; e2, e1, attrs).
>> 9. Constraint 56 should be: IF hadMember(c,e) and
>> 'prov:EmptyCollection' ∈ typeOf(c) THEN INVALID.
>>
>> * PROV-O
>>
>> PROV Ontology contains several axioms for inferencing but it does not
>> cover all the inferences described in the PROV constraints document
>> even though these inferences can be encoded in OWL in a
>> straightforward way. We think these inferences are useful not just for
>> validation but also for querying PROV documents. For this reason, we
>> believe these inferences should be included in PROV-O.
>>
>> Here are some example OWL axioms encoding some of the inferences from
>> PROV constraints document:
>>
>> # Inference 16 (alternate-reflexive)
>> # IF entity(e) THEN alternateOf(e,e).
>>
>> :Entity
>>    rdfs:subClassOf [
>>        a owl:Restriction ;
>>        owl:hasSelf true ;
>>        owl:onProperty :alternateOf
>>    ] .
>>
>> # Inference 17 (alternate-transitive)
>> # IF alternateOf(e1,e2) and alternateOf(e2,e3) THEN alternateOf(e1,e3).
>>
>> :alternateOf a owl:TransitiveProperty .
>>
>> # Inference 18 (alternate-symmetric)
>> # IF alternateOf(e1,e2) THEN alternateOf(e2,e1).
>>
>> :alternateOf a owl:SymmetricProperty .
>>
>> # Inference 19 (specialization-transitive)
>> # IF specializationOf(e1,e2) and specializationOf(e2,e3) THEN
>> specializationOf(e1,e3).
>>
>> :specializationOf a owl:TransitiveProperty .
>>
>> # Inference 20 (specialization-alternate-inference)
>> # IF specializationOf(e1,e2) THEN alternateOf(e1,e2).
>>
>> :specializationOf rdfs:subPropertyOf owl:TransitiveProperty .
>>
>> * PROV-CONSTRAINTS Test Cases
>>
>> We appreciate as implementers the PROV-Constraints test suite. We
>> would like to see test suites for the other operational parts of PROV,
>> in particular for testing inferences separate from validation. This
>> request arises from our general belief that interoperability with a
>> formal spec is typically less high than interoperability with a formal
>> spec *and* an executable test suite. Test suites are invaluable for
>> implementations. Further, while we would like to see the test suites
>> be made normative parts of PROV (since that gives a nice algorithm for
>> resolving disagreements between spec test and test suite (i.e., tie
>> goes to the test suite)), we would prefer non-normative test suites to
>> no test suites at all.
>>
>> We identified the following issues in the following RDF test cases:
>>
>> * prov-o-class-Invalidation-PASS.ttl: At line 37, there are repeated
>> semi-colons ‘;;’ which is invalid according to the Turtle grammar
>> (neither [2] nor [3] seems to allow this).
>> * prov-o-class-Collection-PASS.ttl: Invalid xsd:dateTime literals
>> missing minutes and timezone identifier.
>> * prov-o-property-hadMember-PASS.ttl: Invalid xsd:dateTime literals
>> missing minutes and timezone identifier.
>> * ordering-association2-PASS-c47.ttl: This test is marked PASS but it
>> is inconsistent because the individual ex:ag is an instance of
>> disjoint classes prov:Entity and prov:Activity.
>> * prov-o-property-qualifiedCommunication-PASS.ttl: This test is marked
>> PASS but it is inconsistent because the individual
>> :writing-celebrity-gossip is an instance of prov:Activity but uses the
>> property prov:wasAttributedTo whose domain is the disjoint class
>> prov:Entity. Same argument is also true for the individual
>> :voicemail-interception.
>> * prov-o-property-qualifiedRevision-PASS.ttl: This test is marked PASS
>> but it is inconsistent because the individual :draft2 is an instance
>> of prov:Entity but uses the property prov:wasAssociatedWith whose
>> domain is the disjoint class prov:Activity.
>>
>> We think following tests should not have been included in
>> rdf-tests.txt because the invalid PROV-N constructs cannot be
>> expressed in RDF and thus their RDF representation is valid:
>>            unification-association-f4-FAIL-c23.ttl
>>            unification-association-f5-FAIL-c23.ttl
>>            unification-derivation-f1-FAIL-c23.ttl
>>            unification-derivation-f2-FAIL-c23.ttl
>>            unification-derivation-f3-FAIL-c23.ttl
>>            unification-derivation-f4-FAIL-c23.ttl
>>
>> Best,
>> Evren
>>
>> [1] http://stardog.com/
>> [2] http://www.w3.org/TeamSubmission/turtle/#sec-grammar
>> [3] http://www.w3.org/TR/turtle/#sec-grammar
>>
>>
>> --
>> Evren Sirin
>> CTO
>> Clark & Parsia, LLC
>> http://clarkparsia.org
>>
>
>
>
> --
> --
> Dr. Paul Groth (p.t.groth@vu.nl)
> http://www.few.vu.nl/~pgroth/
> Assistant Professor
> - Knowledge Representation & Reasoning Group |
>   Artificial Intelligence Section | Department of Computer Science
> - The Network Institute
> VU University Amsterdam
Received on Wednesday, 23 January 2013 16:24:56 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Wednesday, 23 January 2013 16:24:56 GMT