On 5/25/11 12:12 PM, John F. Sowa wrote:
> I agree with your point about LOD. But there is nothing in the
> Semantic Web that helps. (01)
Yes, assuming (as I do) you see Linked Data Graphs and RDF as being
distinct :-) (02)
Now in the case of Watson though, the Linked Datasets where in RDF form
(a Semantic Web Project output) prior to ingestion, as part of Watson
learning process etc.. (03)
> Watson just puts the URIs into their
> UIMA notation, stores them in a relational database (DB2),
> and uses Prolog for analyzing and reasoning about Jeopardy
> questions. (04)
Yes, which is also why I really hope Watson has laid the foundation for
some serious future reasoning Olympics for machines aimed as business
intelligence and general knowledge oriented algorithms. Cray, HP, Intel,
and others would ideally be part of aforementioned event. Humans assume
the role the role Race Car Drivers in this future competition :-) (05)
>> But, and this is really really important, it leveraged Linked Datasets
>> such as DBpedia and others from the Linked Open Data cloud.
>> Structured Linked Data, Machine Learning, and Natural Languages are 3
>> equally critical components of the Watson story. Take one of these
>> components out of the mix and Watson doesn't happen.
> First point: The SW does *nothing* for NLP. They ignored it totally.
> They expected human annotators to insert RDF tags. That might have
> been a useful quick fix in 1998. But today, a good NLP system can
> derive much better information from the raw text than from the
> guesses by typical annotators.
> The SW layer cakes emphasized Unicode, URIs, XML, and RDF as the
> foundations. Google used the URIs, but stored them in JSON.
> Watson also used the URIs, represented them in UIMA, and stored
> them in a relational DB.
> RDF is just a bloated notation that anybody who needs efficiency
> runs away from as fast as possible. Even the W3C replaced it
> with RDFa, whose only similarity to RDF is the first 3 letters.
> As for DBpedia, its hierarchy of terms is abysmal. It's better
> than nothing for some applications, but it really needs to be
> rebuilt with some decent NLP tools.
>> I fantasize about a Machine vs Machine edition of Jeopardy
>> (in the future) that pits algorithms against algorithms where
>> all parties (other players in this realm) have a common structured
>> linked substrate as the knowledge base foundation
> I think that Watson pretty much exhausted what we can learn from
> the Jeopardy game. The questions are too short and the format
> depends too heavily on the speed of hitting the buzzer.
> Governments, businesses, and law firms have huge volumes of NL texts.
> There are far more important challenges in analyzing those texts
> and doing the kinds of tasks that people will pay for.
> For more discussion, see http://www.jfsowa.com/pubs/futures.pdf
> Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
> Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
> Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
> Shared Files: http://ontolog.cim3.net/file/
> Community Wiki: http://ontolog.cim3.net/wiki/
> To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
Twitter/Identi.ca: kidehen (09)
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J (010)