W3C home > Mailing lists > Public > public-prov-wg@w3.org > February 2012

Comments on PROV-O

From: Cresswell, Stephen <stephen.cresswell@tso.co.uk>
Date: Thu, 23 Feb 2012 13:40:47 -0000
Message-ID: <F22D0BFCDD4DDC44B92C4E24D751CB93EA7191@W3EXC017023.theso.co.uk>
To: <public-prov-wg@w3.org>

These are comments on PROV-O as I found it on Wednesday (22 Feb 2012).

Generally, I think PROVO-O is looking usable and compatible with PROV-DM
wd3.  The PROV-RDF mapping is really helpful for understanding how
PROVO-O is intended to be used.

- Naturalness of RDF.
I'm a bit scared to see a single record in the PROV-ASE being mapped to
>10 RDF triples, especially if the record was only stating a simple
binary relationship. However, if we're allowed to skip the qualified
involvements when we don't need them and just use the direct properties,
then we could often be using just one triple.  We are allowed to do
that, aren't we?  Also, there is hopefully nothing stopping people from
using their own domain-specific subclasses and subproperties.

- I don't see any mismatches from PROV-DM that aren't already flagged or
under discussion.  I'm still looking.

- At the moment, many of the properties are not defined with all their
characteristics (e.g. prov:tracedTo and prov:specializationOf should be
transitive, prov:alternateOf should be symmetric).  Presumably that's
because the priority has been to get the hierarchies of classes and
properties settled.  It would be really nice to see some of the many
things that could be done to make the ontology come alive and enable
some helpful inferences.  I think that many of the definitions,
constraints and defined inferences of prov-dm can be expressed quite
directly in OWL-RL using property chain axioms, and I really hope that
the plan is to do that - e.g. 
- direct properties (e.g. prov:used, prov:wasGeneratedBy, ...) should be
inferable from the corresponding Involvements.
- wasInformedBy should be inferable from property chains (used,
wasGeneratedBy)
- tracedTo should be inferable using all various paths defined in
prov-dm.
  
Some of the recent changes may be unhelpful for that. By consistently
using properties with generic names (prov:qualified, prov:entity) to
link Involvements, it will be impossible to define the property chains
that would enable the direct properties (prov:used, prov:wasGeneratedBy,
...) to be inferred from them.  To do that, I think there need to be
subproperties which are unique to the different subclasses of
Involvement.  It appears that these have only just been removed.

For example, I don't believe it's possible define the ontology so that
it's possible to infer this...
  :a1 prov:used :e1 .
from this...
  :a1 prov:qualified :u1 .
  :u1 prov:entity :e1 .
  :u1 a prov:Usage .
... but it would be possible to do that if the prov:qualified and
prov:entity properties were specialised for linking prov:Used.


- On first reading, I wrongly thought prov:Quotation was a kind of
Entity rather than a kind of EntityInvolvement, and I correspondingly
misread hadQuoterAgent, hadQuotedAgent property as relating an Entity to
an Agent and thought that it ought to be a specialization of
wasAttributedTo.  The way it's actually modeled does match PROV-DM, so
there's nothing wrong here, but it might be worth considering renaming
prov:Quotation to help avoid the tempting misreading.

Stephen Cresswell

***********************************************************************************************
This email, including any attachment, is confidential and may be legally privileged.  If you are not the intended recipient or if you have received this email in error, please inform the sender immediately by reply and delete all copies from your system. Do not retain, copy, disclose, distribute or otherwise use any of its contents.  

Whilst we have taken reasonable precautions to ensure that this email has been swept for computer viruses, we cannot guarantee that this email does not contain such material and we therefore advise you to carry out your own virus checks. We do not accept liability for any damage or losses sustained as a result of such material.

Please note that incoming and outgoing email communications passing through our IT systems may be monitored and/or intercepted by us solely to determine whether the content is business related and compliant with company standards.
***********************************************************************************************

The Stationery Office Limited is registered in England No. 3049649 at 10 Eastbourne Terrace, London, W2 6LG
Received on Thursday, 23 February 2012 13:41:56 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 26 April 2012 13:06:56 GMT