[Kim-discussion] insert new instances into KIM

[Marin Nozhchev]

Dear Fangkai,

I can give you a detailed answer of the particular question as it 
tackles an important issue . I hope it's not too late.

In RDF, you can have multiple labels (rdfs:label) for a single entity 
like Tokyo, but Tokyo must have a single URI. However, it's impossible 
to make sure that all RDF vocabularies published in the world use the 
same URIs. So you have three options:

1. Preprocess the smaller RDF file, replacing the different URIs of an 
entity with a single one. This will require the most initial effort but 
there will be no downsides afterwards.

2. Add owl:sameAs statements that link the equivalent entities, eg. 
wkb:Tokyo owl:sameAs other:Tokyo . This will make them equivalent at 
database level - that is - the RDF database behind KIM will merge the 
data (e.g. companies located in) about both URIs together. The database 
will also return both URIs as result to queries that match the data 
about any one of them. In practice, however, this leads to both 
ambiguity and increased load on the database. The theoretical behavior 
is undefined in some corner cases and is generally difficult to 
implement by the database. KIM 3 with BigOWLIM 3.3 as RDF database has 
improved support for owl:sameAs. We haven't tested sameAs support in KIM 
2.4 with SwiftOWLIM 2.0 DB, so I can't promise anything there. Learn 
more about sameAs here 
http://journal.webscience.org/403/2/websci10_submission_123.pdf

3. Add semantic metadata in the RDF that explains that the two concepts 
can be used interchangeably. The SKOS mapping property skos:exactMatch 
is the best candidate e.g. wkb:Tokyo skos:exactMatch other:Tokyo (full 
URI of skos:exactMatch : http://www.w3.org/2004/02/skos/core#exactMatch) 
. skos:exactMatch doesn't mean anything to the RDF database - the 
triples will not be interpreted differently than any other data that 
refers to the two Tokyo URIs. In this case, the client application - KIM 
or your custom client - should take this metadata into account. KIM 2.4 
doesn't support SKOS and KIM 3.0 (available now!) has limited support 
for SKOS semantics. Technically, you can alternatively add schema 
statements and inference rules to implement skos:exactMatch semantic 
directly in the database, but in that case you are better off just using 
owl:sameAs as the same problems in corner cases will pop up again. Learn 
more about SKOS here - 
http://www.w3.org/TR/2009/NOTE-skos-primer-20090818/ and see the 
difference between owl:sameAs and skos:exactMatch here - 
http://www.w3.org/TR/2009/REC-skos-reference-20090818/#L4858 .

As you can see, only option 1 is a complete solution. Using the other 
options require that you both upgrade KIM and read the reference 
material so you know what to expect.

Best regards,
Marin Nozhchev

On 21.06.2010 23:12 ч., Yang Fangkai wrote:
> Dear List,
>
>         I have a questinon about inserting new instances into KIM world
> KB. I have a RDF file, and use toolRDFImport.cmd to import it. All the
> entities can be successfully imported. But my question is how to
> establish a connection between synonyms. For example, I import triples
> describing a company which locates in "Tokyo". In the imported RDF
> file, this Tokyo has its unique URI which is different from that in
> wkb.nt. Is there anyway to automatically align these synonyms?
>
> Thank you!
>
> Fangkai
> _______________________________________________
> Kim-discussion mailing list
> Kim-discussion at ontotext.com
> http://ontotext.com/mailman/listinfo/kim-discussion
>