[Top] [All Lists]

Re: [ontolog-forum] API4KB and diverse ontologies

To: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Ray Martin <marsaviator@xxxxxxxxx>
Date: Sat, 29 Jun 2013 18:27:17 -0400
Message-id: <CAAMmmDi8B464ULoNyH+03bhMJ6A1p=HD0g5R-Dfwo3TgBKeA9w@xxxxxxxxxxxxxx>

On Sat, Jun 29, 2013 at 4:29 PM, John F Sowa <sowa@xxxxxxxxxxx> wrote:

Thanks for going into more detail about Apache Stanbol.  I followed
some of those links to explore the various components, their design,
and their APIs.

> As one attempts to produce a system, one realizes that one will
> be thousands of years old if one produces all components by one's
> lonesome.  So, one attempts to use ideas and accomplishments of others
> - Stanbol, API4KB, etc, etc

I have no quarrels with that point.  But the main question is what
kind of platform do you intend to build on?

I am in a quandary of study and confusion at this time. But, i am thinking that Stanbol has some pieces of interest - maybe in the area of 'harvesting' knowledge from the web.

If all you have are APIs, you have some specifications, but no
implementations.  Stanbol is a complete system that implemented
tools and components that support many APIs at different levels.
Unless you have a large team of implementers, you need *software*,
not a bunch of spec's.

The front page of the Stanbol site said that their results are
exported in three very limited formats:  RDF, JSON, and JSON-LD.
But when you dig into their system, they have many different
APIs for linking many different reasoning components, most of
which are developed by other projects.  For each of those
components, Stanbol has to use APIs that are provided by
those components.

i think i like the Stanbol manner of including other projects. 

So that might be a useful base to start from.  But there are
many other useful bases.  The biggest of all is Cyc, which has
over a person-millennia of work behind it.  But I doubt that
many, if any, of their APIs are shared with Stanbol or API4KB.

My current opinion is that i want a set of agents that access services.  A 'monitor agent' says, "hey, my human wants more information on subject, X and Y, but doesn't seem interested in Z.  Another agent that deals with subject X, says X is my forte, so let me check with Cyc and Watson for elements p, q, and r.  So, my current theory was to use FIPA agents (JADE?) to bring together the offerings of various services.  But, then you mentioned Elephant as a glue logic. Now i need to reconsider methodologies.  But, i also need to dig deeper to prepare a set of use cases so that i can better decide "Elephant or agents or ..."  As agents drop off their findings (from Linked Data, Cyc, web page scaping, wiki scraping, etc), a Stanbol like thing maintains a CMS of facts. But, does a CMS really have the appropriate retention elements that ideally assist human cognition? Need visualizations to allow the human's preattentiveness.  Need to assist the human mind maps and images.

I am working on http://www.jfsowa.com/talks/goal.pdf , slide 16 to hold a frame of reference, slide 40 to think about what a human needs from machinery to be assistive, slide 44 for same reason - then from there i am working on up to 158. i am using your pubs to perform additional searches on the subjects represented.  All of this to enable decision of what kind of machine to build for cognitive assistance to the human.

Have included http://www.jfsowa.com/pubs/arch.htm - trying to understand figure 6 and surrounding information.

i am really liking:
http://www.brain-mind-media.org/archive/150 , The Role of Conciousness in Memory - i am looking further at the IDA model to which the article refers.

Any way, the more I read, the more I understand how far I have to go.  For proof to one's self that cognition of a subject has occurred, one needs to take action.  In the near future, i should attempt a course of action to locate the shortcomings in what i think i know.  But, before i do that i am going to go back and look closely at some of your architecture discussions. Folks who have been there, done that will keep me from landing on the moon when i was striving for Mars.

Another very large base is IBM's Watson, which uses the UIMA
architecture, representations, and APIs.  IBM is licensing Watson
for various purposes, for which it might be very useful.  But I
doubt that it would share many APIs with Stanbol or with API4KB.

A historically important system, whose basic design is still sound,
is Ontolingua.  Following is a report by Tom Gruber from 1993:

    A translation approach to portable ontology specifications

I was just reading that link this morning.  I read, now i need to go back and absorb.
The translation approach was the consensus of a large group
of AI experts who gathered together in a loose collaboration
called SRKB (Shared Reusable Knowledge Bases).  The founding
meeting was in 1991, and several projects came out of it:
KIF, KQML, and Ontolingua among them.  For a summary of the
translation approach, look at Fig 1 (p. 7) and Fig 2 (p. 16)
of Tom G's article.

Other projects were strongly influenced by SRKB, including UMLS
and Protege.  Another influence was the collaboration between
KIF (Genesereth & Fikes), CGs (Harry Delugach, me, and several
others), and several different NCITS committees. That eventually
led to the Common Logic standard in 2007.

The translation approach was also the foundation for SWeLL -- the
very general and expressive logic in Tim B-L's proposal of Feb. 2000:


In that document, search for every occurrence of 'swell'.  That
leads to the original Semantic Web "layer cake", which I believe is
an *improvement* on its successors.  Just below that layer cake,

Tim B-L
> The Semantic Web, then, will serve as an interchange "bus" for on-line data.
> In effect, it will allow any web software to read and manipulate data published
> by any other web software. SWeLL will enable this interchange. The pairing of
> simple, predictable, reliable systems with complex, unpredictable, heuristic
> systems is one of the novel possibilities opened by the Semantic Web.

This is a clear, concise statement of the translation approach. It is
consistent with the consensus of the SRKB project, which was implemented
in Ontolingua and other projects.

Pat Hayes participated in the SRKB project, the Common Logic project,
and the early work on the SW.  He worked with Guha to produce the LBase
specification, which is compatible with Tim B-L's proposal for SWeLL.

If LBase had been adopted in 2003 along the lines that Tim B-L had
proposed for SWeLL, we wouldn't have any of these debates today.


I am still working on your latter paragraphs.  Thanx for sharing,

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    (01)

<Prev in Thread] Current Thread [Next in Thread>