Frank, (01)
Your slides present a good, crisp overview of the main points
covered in the communique. A copy of them should be stored on
the wiki with cross references to and from the communique. (02)
I would also like to comment on slide 29: (03)
Knowledge Representation Languages (04)
- Resource Description Framework (RDF)
- OWL-DL (description logic)
- OWL-Full (first order logic)
- Common Logic (first order logic)
- OBO (frame based) (05)
That is already a large number of notations, all of which could be
mapped into Common Logic as the common base language. But I doubt
that the users of those other languages will soon stop using their
favorite notations and switch to CL. I also doubt that the number
of KR languages can be limited to just that list of five. (06)
That point raises two further questions: (07)
1. How can we enable computer systems that use different KR
languages to interoperate? (08)
2. How can we enable humans who know one KR language to read and
understand specifications written in a different KR language? (09)
Common Logic was designed in response to questions such as #1, but
there still remains a lot of work to be done on the very many thorny
issues involved. (010)
But question #2 is at least as important, since it is essential
for the designers, developers, and users of interoperating systems
to understand what how those systems communicate, what they say in
their messages, and whether what they are saying is accurate. (011)
For a long time, I have claimed that we must have controlled
natural languages (for starters, some version of controlled English)
as an auxiliary notation for *all* knowledge representation languages. (012)
Adam Pease was initially skeptical about that idea, but he eventually
adopted it for SUMO, and they found it very useful. Other groups,
such as Cyc, spent a lot of effort to make a more readable version
of their notation, although it was not really a controlled English
that could be read by a person who did not know a lot about Cyc. (013)
Trying to design a single syntax for controlled English that maps
to each of those five languages (plus any others that might be
designed) would be a major challenge. But that is not really
necessary because people who can read English can read several
variants of controlled English with little or no training. (014)
Therefore, my recommendation is the following: (015)
1. For every KR language used in the OOR, there shall be a version
of controlled English that can be automatically mapped to that
language (and possibly from that language as well, but mapping
to the KRL is the minimum requirement). (016)
2. For every statement in every ontology in the OOR, there shall
be two versions:
a) A formal statement in one of the approved KR languages.
b) A logically equivalent statement in controlled English. (017)
3. The equivalence of the KRL statement (2a) and the controlled
English statement (2b) shall be automatically verified by
appropriate software. (018)
For point #3, the simplest way to verify equivalence is to generate
the KRL by translating from the controlled English. But it would
also be possible to translate the KRL to CE. Another option is to
use an optimizer that takes KRL that had been translated from CE
and performs further translations to make it more efficient. (019)
For some examples of controlled English, its relationship to ordinary
English, and its mapping to Common Logic, see the following slides: (020)
http://www.jfsowa.com/talks/cl_sowa.pdf (021)
John Sowa (022)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/oor-forum/
Subscribe: mailto:oor-forum-join@xxxxxxxxxxxxxxxx
Config/Unsubscribe: http://ontolog.cim3.net/mailman/listinfo/oor-forum/
Shared Files: http://ontolog.cim3.net/file/work/OOR/
Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?OpenOntologyRepository (023)
|