Re: Afghanistan War Diary as topic map in Maiana

Aki,

On Tue, 2010-11-02 at 22:15 +0200, Aki Kivelä wrote:
> 1.11.2010 21:08, Patrick Durusau kirjoitti:
> > Aki,
> >
> > Thanks for the work on the map and the report on the extractors!
> >
> > Do any of the extraction services add properties to what we would call
> > subjects?
> 
> Hi Patrick
> 
> If you consider a subject something that has a subject identifier then 
> almost everything given by extraction services are subjects. For 
> example, keywords provided by the Alchemy web service are subjects that 
> represent keywords provided by the Alchemy. This may sound a tautology, 
> but I think it is important that a subject reified by an extractor is 
> not even trying to pretend something more it really is.
> 

Well, actually I have a fairly broad notion of what a subject can be and
it doesn't involve the presence or absence of a subject identifier. 

True, having a subject identifier or a published and resolvable subject
indicator can be important for some purposes but neither one determines
(for me at any rate), the presence of a subject. 

For example, assume the war diaries include a report of an incident at
some identifiable location in Kabul. The location is a subject, that it
exists in an association with Kabul and the incident in question are
both subjects, that the people involved (who are also subject) were in
an incident at that location is also a subject (association), and so on.

In creating a topic map from that raw data I might or might not choose
to recognize all those subjects and/or to assign them identifiers, etc.

If I wanted to study patterns of incidents at particular locations, then
creating associations to represent those subjects, along with all the
other subjects that would require, could be quite important. 

> If you consider a subject something that has a published and resolvable 
> subject indicator then I have to disappoint you. I haven't really 
> investigated all extracted topic maps yet but it looks like only some 
> entities provided by Alchemy have well known identifiers referring 
> geonames, dbpedia, freebase, yago, umbel, opencyc etc. For example, 
> Alchemy entity CHINA with an extractor generated subject identifier 
> http://www.alchemyapi.com/api/entity/CHINA has been labeled same as
> 
> http://dbpedia.org/resource/People's_Republic_of_China
> http://mpii.de/yago/resource/People's_Republic_of_China
> http://rdf.freebase.com/ns/guid.9202a8c04000641f8000000000c01763
> 
> "Labeled same as" means here that the entity topic of CHINA has been 
> associated with three topics, each identified by one subject identifier 
> written above, and association type is a home-grown "sameAs". A kind of 
> home-grown linked data, you see. Another question is, how subject CHINA 
> appears from AWD reports.
> 

No apologies necessary! The linked data folks are finding out that
quality semantic data is a fairly rare commodity. 

Of course associations could use fairly coarse role players, like China,
U.S., India but that probably isn't useful for all cases. ;-)

> Did I manage to answer your question?
> 

Yes.

Thanks again!

Patrick

> Kind Regards,
> Aki Kivelä
> Wandora Team
> _______________________________________________
> topicmapmail mailing list
> [email protected]
> http://www.infoloom.com/mailman/listinfo/topicmapmail
_______________________________________________
topicmapmail mailing list
[email protected]
http://www.infoloom.com/mailman/listinfo/topicmapmail