Re: How do RDF and Formal Logic fit together?

[Thomas B. Passin]
> 
> To hook this into the threads on literals, how is your DatExpert supposed to
> know that a given string is to be analyzed as a date, if we cannot make a
> statement about the type of the literal?  The only other ways would seem to
> be textual (or contextual) analysis which are pretty hard, or human
> intervention which is fine but not so Semantic Webish.
> 

The name of a slot or property ("Birth date") should give you an
indication that a date is involved (semantics for human use and machine
exploitation if we tell them to use this in particular cases). 

The type of property value ("date") tells you it should be looked upon
as a date (semantics for machine exploitation). 

But the format, at least in my broader perception of dates, will always
be a string in its physical representation. 

The way the string should be interpreted (logical format?) making it
machine understandable in terms of a date can be added if known (Format:
string = referring to a node | yyyy-mm-dd | DatExpert:absoluteValue |
…).

For computer generated RDF, XTM or whatever you might want to use (and I
hope this will be the case soon for everything newly created in the near
future) in some (most?) cases the format will be known and should be
added (reducing future possible misinterpretation).

Now have a look at what I described as the third phase of the semantic
web (all devices are nodes in the network and the embedded software of
it). The checker services will not use static dictionaries but will use
the dynamic information of the semantic web providing besides spelling
and grammar checking also a semantic checking. Contextual information
from the device itself and the operator will also be integrated.
When I write or speak about "Cretaceous" behind my back this "word" will
be linked to a node representing this geological period and through this
with all the info already known about it (including begin and end).

But at the same time we do have a tremendous amount of legacy
information that doesn't have the machine exploitable form and that
should be "annotated" with it. We must provide high quality textual (and
contextual) analysis and that must be done without human intervention.

 
Ronald Poell
Consultant Knowledge Management
Netherlands Organization for Applied Scientific Research (TNO)

Oude Waalsdorperweg 63
P.O. Box 96864
2509 JG The Hague
The Netherlands

T +31 70 374 02 00
F +31 70 374 06 52

http://www.tno.nl

email poell@fel.tno.nl

Received on Sunday, 7 October 2001 02:33:22 UTC