Krzysztof and Kingsley, (01)
I mostly agree with your points, but I'd like to add some comments. (02)
KJ
> To start even one step earlier there are no rivers, lakes, or forests in
> the physical world. Their conceptualization is based on cognition and
> especially social convention. All these types can only be defined in a
> local context (microsenses; to use your terminology), i.e., taking space
> and time into account. (03)
The term 'microsense' was coined by the linguist Alan Cruse to describe
the open-ended number of fine distinctions and variations that are
possible for almost any word in any natural language. Rivers, lakes,
and forests are typical examples, where you can't define any clear
boundary between river, stream, brook, creek, etc. Usually the
distinction is based on some human scale: if you need a bridge or
a boat to cross, it's a river; but if you can ford it, it's a stream. (04)
KJ
> Meaning only emerges within a context by situated simulation... (05)
Yes, and a context is always dependent on some reason or purpose
for making the distinction. A related term is 'situation', which
is delimited by the range of perception of the agents (people,
animals, and robots) in the situation. (06)
KJ
> I believe that with a growing number of facts in the
> Linked Data cloud the semantic integration problem will become
> much more obvious and work on semantic translation, alignment,
> induction, similarity, etc will become more prominent. (07)
I agree. But being more obvious does not automatically lead to
a more obvious solution. The critical issue is still the reason
or purpose for making a distinction. With more people involved,
you get an increase in the number and kinds of purposes. (08)
KJ
> The original documents provided the creation context which is necessary to
> to determine which microsense has to be used to understand a specific term.
> In this sense, Linked Data de-contextualizes data. (09)
Actually, I would say that it just adds more contexts -- provided that
you save the original source data. (010)
KJ
> I am very curious to see which role the 'ontological layer' will play
> in the future of Linked Data and hope these ontologies will be derived
> (and personalized) our of real data, i.e., bottom-up. (011)
I agree. Ferrucci made the point that an a priori annotation is usually
restricted to somebody's preconceptions. But they found that the best
annotations are done from the point of view of the person asking
a question -- which may be very different from the author's POV. (012)
I believe that is the flaw in trying to build huge ontologies or
knowledge bases like Cyc. They're defined top-down from a point of
view that may have no relationship to the question or application. (013)
KI
> DBpedia, and many datasets from the LOD cloud provided great sources of
> facts to Watson. (014)
I fully support the idea of using whatever resources are available
-- both unstructured and highly structured. At VivoMind, we use
anything and everything we can find, and we make it available to
any process that might need it. Since retrieval time is logarithmic,
it doesn't cost much to add more data -- provided that you have
enough disks to store it. (015)
KI
> Watson is a demonstration of what can be achieved when you combine NLP
> and Deductive DBMS technology -- where Linked Data and lots of other
> Semantic Web Project outputs contribute. (016)
But I would add other points that Ferrucci emphasized: machine
learning, the derivation and testing of multiple hypotheses,
statistic, and heuristics. I always emphasize that deduction is
only 25% of Peirce's Cycle of Pragmatism; induction, abduction,
and testing are the other 75%. (017)
I believe the greatest strength of Watson is their combined use
of multiple paradigms. That is the theme of the following paper: (018)
http://www.jfsowa.com/talks/paradigm.pdf
Two Paradigms are Better than One,
And Multiple Paradigms are Even Better (019)
John (020)
_________________________________________________________________
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
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (021)
|