Pedantic Web Challenge 2011
Hogan, Aidan
Welcome to the inaugural Pedantic Web Challenge.
This years question is:
"Why is the description of the W3C organisation (as identified by, e.g.,
[1]) inconsistent on the Web of Data?"
(Just to confirm that the W3C's description *is* inconsistent.)
The winner will be the first answer on this list. Bonus points for
saying how you found it, or contacting the folks involved through this
list. The prize will be bragging rights to anyone who will listen, or at
least hear.
(Feel free to pass word on. I'm curious to see if other people are
coming across these sorts of problems in their apps, or how others might
go about finding such problems.)
If we don't get any entrants, I'll give my answer next Friday.
Happy hunting!
Aidan
Antoine Zimmermann
Or a weaker notion of inconsistency?
Le 06/01/2011 21:27, Hogan, Aidan a �crit :
Hogan, Aidan
> Or a weaker notion of inconsistency?
Yep, with respect to OWL (1) semantics... or, if you prefer, lets say
with respect to applying the closure of OWL 2 RL/RDF rules over the data
[1].
Antoine Zimmermann
On http://dig.csail.mit.edu/People/RRS/about, there are RDFa annotations
saying that http://www.w3.org/data#W3C has a foaf:homepage which is
http://www.w3.org/. See the following snippet:
"""He is a member of the staff of the
<span about="http://www.w3.org/data#W3C">
<span property="foaf:name">World Wide Web Consortium</span>
[<a rel="foaf:homepage" href="http://www.w3.org/"
><span property="s:label">W3C</span></a>]
<link rel="con:publicHomePage" href="http://www.w3.org/" />
</span>"""
Moreover, on http://identi.ca/w3c/foaf, it is said that
http://identi.ca/user/48404 has foaf:homepage http://www.w3.org/. Since
foaf:homepage is an inverse functional property, I deduce that
http://identi.ca/user/48404 and http://www.w3.org/data#W3C are the same
thing.
Now, still in http://identi.ca/w3c/foaf, I see that
http://identi.ca/user/48404 foaf:knows a lot of people. Since the domain
of foaf:knows is foaf:Person, I can deduce that
http://www.w3.org/data#W3C is of type foaf:Person.
Now, if I dereference http://www.w3.org/data#W3C, I see that this entity
is a foaf:Organization. FOAF says that foaf:Organization and foaf:Person
are disjoint. So, we end up with an entity, namely
http://www.w3.org/data#W3C, which is a member of two disjoint classes
(viz., foaf:Person and foaf:Organization) which makes an inconsistency.
Now, the culprit in this horrible reasoning mess is the identi.ca
dataset. But identi.ca is not really responsible of the issue. They have
members, which are supposed to be people registering to their service.
If a registered member pretends to be an organisation, we can only blame
the registered member, not the service.
So, the next step if I want to claim the bonus points would be to
contact the member who registered to identi.ca with the nick w3c
(http://identi.ca/w3c). I don't see an email address on that page, so I
won't try contact this person. Moreover, there may be other datasets
with the same kind of mistakes, so my efforts would be vain...
AZ.
Le 06/01/2011 21:27, Hogan, Aidan a �crit :
Antoine Zimmermann
to say that I simply used SWSE [1] for solving this riddle. Following
SWSE's results and links, and with a bit of intuition, I found the
documents mentioned in my reply. There is probably (so far) no better
tool than SWSE to solve this kind of challenge.
(PS: I was not bribed to say this !)
AZ.
2011/1/6 Antoine Zimmermann <antoine.z...@gmail.com>:
> Le 06/01/2011 21:27, Hogan, Aidan a écrit :
>>
>> Hey folks,
>>
>> Welcome to the inaugural Pedantic Web Challenge.
>>
>> This years question is:
>>
>> "Why is the description of the W3C organisation (as identified by, e.g.,
>> [1]) inconsistent on the Web of Data?"
>>
>> (Just to confirm that the W3C's description *is* inconsistent.)
>>
>> The winner will be the first answer on this list. Bonus points for
>> saying how you found it, or contacting the folks involved through this
>> list. The prize will be bragging rights to anyone who will listen, or at
>> least hear.
>>
>> (Feel free to pass word on. I'm curious to see if other people are
>> coming across these sorts of problems in their apps, or how others might
>> go about finding such problems.)
>>
>> If we don't get any entrants, I'll give my answer next Friday.
>>
>> Happy hunting!
>> Aidan
>>
>> [1] http://www.w3.org/data#W3C
>>
>>
>
>
--
--AZ
Hogan, Aidan
> Oops, it seems Aidan did not want me to enter the contest, but I have
to
> say that I simply used SWSE [1] for solving this riddle. Following
SWSE's > results and links, and with a bit of intuition, I found the
documents
> mentioned in my reply. There is probably (so far) no better tool than
SWSE > to solve this kind of challenge.
People might think it's a fix. You're not helping this impression! :)
> That's an easy challenge :)
We'll have to set a harder one next year.
You win the Pedantic Web Challenge 2011 bragging rights! Congrats! ;)
> Now, the culprit in this horrible reasoning mess is the identi.ca
> dataset. But identi.ca is not really responsible of the issue. They
have
> members, which are supposed to be people registering to their service.
> If a registered member pretends to be an organisation, we can only
blame
> the registered member, not the service.
> So, the next step if I want to claim the bonus points would be to
> contact the member who registered to identi.ca with the nick w3c
> (http://identi.ca/w3c). I don't see an email address on that page, so
I
> won't try contact this person. Moreover, there may be other datasets
> with the same kind of mistakes, so my efforts would be vain...
Users cannot be culprits. The customer is always right! Blaming users is
a fruitless (although often enjoyable) exercise.
This is a complex issue:
- Various parties on the Web (including W3C themselves) correctly say
that the W3C is a foaf:Organisation with foaf:homepage
"http://www.w3.org/"
- The W3C sets up an identi.ca profile, and fills the same homepage
into the signup form
- identic.ca uses foaf:knows to represent social networks (as well as
sioc:follows)
- FOAF states that the domain/range of foaf:knows is foaf:Person and
that foaf:Person and foaf:Organisation are disjoint.
<IMHO>
I find it hard to point the finger, but I'm increasingly of the opinion
that rdfs:domain should be brought round back and shot, and rdfs:range
should only be allowed for datatype literals. This is a little against
the current wisdom, but you have to question: *what are these axioms
buying us?*
Typically they infer class memberships that are already known: where
they're not already known, they're usually incorrect (I should do some
tests on this). I've seen various instances of this. We already know
that DBPedia was using foaf:img for all kinds of resources when foaf:img
has domain foaf:Person. In fact, with a little rdfs:domain reasoning,
DBPedia's version of the W3C [1] thinks it's a dbo:Person because of the
range of dbo:owner [2] (though it's not inconsistent due to a lack of
disjointness constraints).
These sorts of domain and range axioms are more hassle than they're
worth in my opinion. If you want to model domains and ranges, model them
in the label of the property (foaf:personalImg) or in the prose spec. In
any case, I'm largely against class-specific properties, so a plain
foaf:img with no domain or range would suit me just fine...
</IMHO>
That said, I don't see any clear blame in this case. My inclination for
a fix would be to get FOAF to weaken the foaf:knows domain/range axioms.
Another possibility would be to get identi.ca [3] to remove the
foaf:knows relations and just go with sioc:follows on users.
Thoughts?
Cheers,
Aidan
@Antoine, the second part of your prize is the privilege of contacting
W3C, identi.ca, FOAF and DBPedia to let these guys know. :)
[1] http://dbpedia.org/page/World_Wide_Web_Consortium
[2] http://dbpedia.org/ontology/owner
[3] http://identi.ca/w3c/foaf
Bob Ferris
> That said, I don't see any clear blame in this case. My inclination for
> a fix would be to get FOAF to weaken the foaf:knows domain/range axioms.
> Another possibility would be to get identi.ca [3] to remove the
> foaf:knows relations and just go with sioc:follows on users.
>
> Thoughts?
+1 for using sioc:follows as primary relationship type in identi.ca
profiles for associating other agents
Maybe one can introduce for the foaf:knows relationship type a separate
assignment, since this association should express from my point of view
a more social relationship than following someone or something.
Cheers,
Bob
Andreas Harth
> Do you mean "inconsistent" as in logic, based on, e.g., OWL semantics?
> Or a weaker notion of inconsistency?
I'd assume that the triples involving foaf:homepage and foaf:primaryTopic
are strong contenders.
Pointing out the missing slash of http://www.w3.org would be pretty
pedantic indeed, so I won't.
Best regards,
Andreas.
Richard Cyganiak
On 6 Jan 2011, at 23:43, Hogan, Aidan wrote:
> That said, I don't see any clear blame in this case. My inclination for
> a fix would be to get FOAF to weaken the foaf:knows domain/range axioms.
> Another possibility would be to get identi.ca [3] to remove the
> foaf:knows relations and just go with sioc:follows on users.
>
> Thoughts?
I'd put the blame on identi.ca for using foaf:knows as the relationship between its users.
“foaf:knows - A person known by this person (indicating some level of reciprocated interaction between the parties).” [1]
This property is inappropriate here for two reasons.
First, identi.ca doesn't seem to discourage organizations to have accounts, but foaf:knows assumes that it would only be persons.
Second, following someone on identi.ca does not imply “reciprocated interaction” -- there are probably users that are followed by thousands of other users, and it's very unlikely they have interacted with all of them!
But in all fairness, it's hard to say what would be the Right Thing to do for identi.ca here. Part of the blame lies with FOAF for not providing any less restrictive properties for relating people or agents. Sure there's sioc:follows, but it relates *accounts* rather than *agents*, so it's on a different level of abstraction.
(In the spirit of fixing things rather than just complaining (or rather, in the spirit of fixing things *in addition to* just complaining), I talked to FOAF's Dan Brickley on #swig about the possibility of a new property in FOAF -- see [2] for chatlog, and I took an action to write a blog post on the issue. Oh and Aidan, you don't want to click on [3]!)
Best,
Richard
[1] http://xmlns.com/foaf/spec/#term_knows
[2] http://chatlogs.planetrdf.com/swig/2011-01-07#T12-12-45
[3] http://chatlogs.planetrdf.com/swig/2011-01-07#T12-31-47