Hi Jens,

sorry for resuming this old discussion but I'm working on this right  
now (when I started the discussion it was in an evaluation state).

Il giorno 17/lug/2009, alle ore 08.22, Jens Lehmann ha scritto:

>
> Hello,
>
> Piero Molino schrieb:
>> Hi Jens,
>>
> [...]
>
>>>> Is there some kind of limitation i'm not aware of that can
>>>> stop me doing what i described?
>>>
>>> In your description, you assume that there is one class for each  
>>> object.
>>> In general, an object can be instance of several classes. In  
>>> particular,
>>> it can also belong to several "most specific" classes. However, this
>>> does seem to be rare in the DBpedia ontology (and you can  
>>> generalise the
>>> above description to this case).
>>
>> Ok i get it. Now for example let's take:
>>
>> http://dbpedia.org/page/Bari
>>
>> (my home town). the rdf:type property (wich i'm assuming is the one
>> useful for the maping) gives back:
>>
>> rdf:type <http://www.w3.org/1999/02/22-rdf-syntax-ns#type>   
>>
>>    * dbpedia-owl:Place <http://dbpedia.org/ontology/Place>
>>    * dbpedia-owl:Area <http://dbpedia.org/ontology/Area>
>>    * dbpedia-owl:Resource <http://dbpedia.org/ontology/Resource>
>>    * dbpedia-owl:PopulatedPlace
>>      <http://dbpedia.org/ontology/PopulatedPlace>
>>    * http://dbpedia.org/class/yago/AncientGreekCities
>>    * http://dbpedia.org/class/yago/CitiesAndTownsInApulia
>>    * http://dbpedia.org/class/yago/CoastalCitiesAndTownsInItaly
>>    * http://dbpedia.org/class/yago/PortCitiesInItaly
>>
>>
>> Googling yago i've found it's an ontology based on wordnet structer
>> (more or less). By the way as you told me the classes are one more
>> specific than another. Is there a way to determinate how "deep" a  
>> class
>> is other than calculating a path to owl:Thing ? I'm asking this  
>> because
>> right now i'm thinking of mapping an instance to one class, maybe the
>> most specific one, by te way i may find come other ways like map to
>> every class and than take the deepest... i don't know i will have to
>> think a bit more about this :)
>
> DBpedia has different class hierarchies (DBpedia ontology, YAGO,
> OpenCyc, Umbel), which you should not mix in your approach. See  
> Section
> 3.2 in our latest DBpedia paper [2] for an overview. The DBpedia
> ontology has the prefix http://dbpedia.org/ontology/.
>
> Since we currently store all types of an entity (Place, Area,
> PopulatedPlace) for an entity and not just the most specific one
> (PopulatedPlace), you could also calculate the depth by just counting
> the number of classes. This works if there is a single most specific
> class and we keep storing all more general classes in the SPARQL
> endpoint (which might change in the future).

Analyzing the DBpedia ontology I've found that it is somehow  
"lightweight" in respect of the task i want to do on it. To calculate  
a semantic distance maybe YAGO fits better. Also it is based on the  
wordnet structure witch has been used in many projects as a resource  
for calculation of semantic distance. By the way the idea you gave me  
about calculating the distance counting the classes works in the  
DBpedia ontology because every class listed is a subclass of the  
other, but as i am observing from the example i posted before, in the  
YAGO ontlogy it is not like that. I would ask this question on a YAGO  
mailing list but i can't find any reference on this on their website  
so i a m sking here, i hope not o be out of contest.

Another question: ad i want to have a "general" approach, i am finding  
some instances that are somehow different from the others as they  
don't have a rdf:type. For example:
http://dbpedia.org/page/Vampire
Assuming i'm calculating the semantic distance from owl:thing  
caounting the classes in rdf:type, this node would behave differently,  
so how could i do it in this cases?

Thank you in advice.
Regards,

Piero

------------------------------------------------------------------------------
Come build with us! The BlackBerry&reg; Developer Conference in SF, CA
is the only developer event you need to attend this year. Jumpstart your
developing skills, take BlackBerry mobile applications to market and stay 
ahead of the curve. Join us from November 9&#45;12, 2009. Register now&#33;
http://p.sf.net/sfu/devconf
_______________________________________________
Dbpedia-discussion mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/dbpedia-discussion

Reply via email to