W3C home > Mailing lists > Public > semantic-web@w3.org > July 2017

Re: [DBpedia-discussion] Call for Ontology Editor demos for DBpedia

From: Peter F. Patel-Schneider <pfpschneider@gmail.com>
Date: Sat, 15 Jul 2017 08:04:56 -0700
To: Jeff Thompson <jefft0@gmail.com>, Sebastian Samaruga <ssamarug@gmail.com>
Cc: semantic-web at W3C <semantic-web@w3c.org>, DBpedia <Dbpedia-discussion@lists.sourceforge.net>
Message-ID: <4584032d-dea0-954e-132a-e27466ced48a@gmail.com>
On 07/15/2017 12:09 AM, Jeff Thompson wrote:
>> It is always possible to replace classes by properties.
> 
> The mechanism in an ontology for filtering out nonsense is disjointness and
> you need classes for disjointness. For example, the ontology should entail
> that the City class and Government class are disjoint. if an item "Berlin" is
> both an instance of City and Government, then it is nonsense because the item
> is conflating two different senses of Berlin. (Or, equivalently, Berlin has
> two properties whose domains are disjoint classes.)

Even this doesn't need classes (or at least not named classes with explicit
instance links).  Classes can be replaced by boolean properties.  Disjointness
between classes can be replaced by property exclusions, e.g., no object can
have value true for both cityness and governmenthood.

Whether this is good modelling practice is a separate matter.

> The DBpedia community and ultimately the DBpedia organizers need to make a
> fundamental decision: Is DBpedia going to filter out nonsense (or just provide
> searchable database of collected nonsense)? If DBpedia is going to filter out
> nonsense, then it needs disjoint classes.

I wouldn't put this so bluntly.  As well, this is not a simple boolean choice.

DBpedia can go multiple ways.  The current way appears to be along the lines
of accepting information from Wikipedia even when this information doesn't fit
into the DBpedia ontology, e.g., the Green Bay Packers home city is Lambeau
Field, plus adding its own infelicities, e.g., monasteries are buildings.  A
different way to go would be ensure that there are no internal contradictions
in DBpedia at all.  A middle ground might be just to ensure that the DBpedia
ontology matches Wikipedia.

The current DBpedia is not nonsense.  The density of errors is quite low.  I
would certainly like for there to be fewer errors in DBpedia and, in
particular, that Wikipedia infelicities that are identified by using DBpedia
are fixed in Wikipedia.

> - Jeff

peter




> 
> On 2017/07/14 19:34, Peter F. Patel-Schneider wrote:
>>
>> On 07/06/2017 11:56 AM, Sebastian Samaruga wrote:
>>> Question: isn't it possible to 'aggregate' classes of subjects in respect to
>>> the properties / predicates some set of subjects have in common. Example: a
>>> Person class subjects would have 'birthPlace', 'birthDate' and 'name'
>>> properties and an Artist subclass would have those properties of Person plus
>>> 'creatorOf' properties of artworks objects. So a superclass would have a
>>> superset of the properties of a subclass.
>>>
>>> Sorry for my ignorance. Best,
>>> Sebastian.
>>
>> It is always possible to replace classes by properties.  For each class set up
>> a boolean property that signals membership in the class.   This is not a good
>> idea in all cases, though.
>>
>> It is sometimes possible to replace classes by commonalities in other data.
>> For example, person born in the seventeenth century can be those people whose
>> birthdate is in the seventeenth century.   This does depend on the other data
>> being complete in some sense.  Sometimes, however, this is not possible.  What
>> makes an animal a person, for example?  From here we get the notion of natural
>> kinds.
>>
>> So maybe artist can be replaced by those people who have created some artwork,
>> except that then I want to claim that some person is an artist without
>> providing the artwork.   Maybe I can make this claim (exists produced artwork
>> in OWL, for example) but maybe the underlying representation language can't
>> state it (primeness of numbers for example).
>>
>> By the way, superclasses would have a subset of the properties of their
>> subclasses, not a superset.
>>
>> Peter F. Patel-Schneider
>> Nuance Communications
>>
>>
>> ------------------------------------------------------------------------------
>> Check out the vibrant tech community on one of the world's most
>> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
>> _______________________________________________
>> DBpedia-discussion mailing list
>> DBpedia-discussion@lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/dbpedia-discussion
>>
> 
Received on Saturday, 15 July 2017 15:05:28 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 19:50:36 UTC