ontolog-forum
[Top] [All Lists]

Re: [ontolog-forum] Wolfram Alpha

To: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: "John F. Sowa" <sowa@xxxxxxxxxxx>
Date: Mon, 09 Mar 2009 20:40:08 -0500
Message-id: <49B5C4F8.6020608@xxxxxxxxxxx>
Kingsley, Erick, Len, Pat, Jack, Duane, and Adrian,    (01)

KI> Very simple question:  What happens when I don't ask my
 > question in English ?    (02)

Very simple answer:  the current version doesn't understand
any language but English.  But I agree with Erick:    (03)

EA> I think that Wolfram also considered that as well as taking
 > into account shorthand notations...    (04)

Wolfram has addressed the difficult problem of representing
multiple knowledge domains in a formal system of reasoning and
computation.  Mapping their representation of those domains to
and from other languages could be done.    (05)

LY> Any guesses about what is behind this "miracle". According
 > to some beliefs stated on this forum for years - there got to
 > be some "Upper Ontology" involved in this new search engine
 > for it to do what is claimed. Or may be, as suggested by others,
 > - a lattice of theories.    (06)

I don't know the details, but I do know that the Mathematica
system is an outstanding product that is widely used by the
major scientific and engineering R & D institutions around
the world.  From what Wolfram said in his note, I gather that
they have implemented a collection of domain models (as I said
before, the equivalent of an ontology plus a reasoning system).    (07)

I don't know whether they have an upper ontology or a lattice
of theories, but judging from what has worked in other systems
such as Cyc, I would assume that their low-level domain models
are far more important than whatever upper level they may have.
For more information about Wolfram's current products, see the
web site for Mathematica:    (08)

    http://www.wolfram.com/products/mathematica/index.html    (09)

The original Mathematica system of 1988 was built on top of
Prolog.  During the past 20+ years, they developed their own
logic programming system, which incorporates many innovations
that go far beyond plain vanilla Prolog.  The community of
Mathematica users have implemented the axioms for nearly every
major mathematical system on earth.  There are also bindings
from Mathematica to the major programming languages.    (010)

At VivoMind, we use a lot of high-powered mathematics, and we
use Mathematica to test our algorithms.  We invoke Mathematica
from the languages we use (Prolog, Java, and C++) to run the
Mathematica versions interpretively.  After we are satisfied
that the algorithms work, we translate them to lower-level
languages such as C++ for better performance.    (011)

LY> I will venture to say that this is another hyped-up
 > promotion for cash-strapped venture. Remember when Google
 > started there was no hype - it just came and just worked.    (012)

I can assure you that there is no hype in Mathematica.
It works, and they have been profitable for over 20 years.    (013)

PC> Question-answering systems that depend on templates and
 > cannot extract information automatically from text can be
 > very useful and interesting, but probably have little to
 > add to what is already known about NLU.    (014)

JR> Google only brought us a better way to do SEARCH while
 > Wolfram may be bringing us FIND, the real game changer.
 > Some years back ASK Jeeves tried to bring us a question
 > answering capability but didn't quite make it.    (015)

The focus of the Wolfram Alpha project is not on NLU, but
on the domain modeling and reasoning capabilities.  That
gives them a far more robust reasoning component than
systems such as Ask Jeeves.    (016)

JR> I'll bet that Alpha leverages one or more hardware
 > constructs that differ considerably from the stored
 > program computer thusly allowing quick, cheap resolution
 > of combinatorial situations.    (017)

Mathematica runs on Windows, Mac, and Linux on a variety
of hardware platforms.  It's possible that they could derive
better performance from some hardware enhancements, but I
suspect that they are using their current Mathematica tools
to support Wolfram Alpha.    (018)

DN> At XML Global, we had a natural language interface built
 > into our contextual XML search engine in 1998.  We dropped
 > it in favor of an API with structured input and filter queries.    (019)

 From what I know of Mathematica, I'm sure that they have a API
that would support direct access to the reasoning modules from
external programs.    (020)

DN> Of course the biggest Achilles heel was the fact that people
 > who marked up content could make mistakes or worse  would attempt
 > to deceive the system to gain higher search engine rankings.
 > What excites me most about Wolfram Alpha is the potential for
 > reasoning where there are multiple, conflicting bits of data.    (021)

The problem of dealing with multiple conflicting bits of data
is definitely important.  I doubt that the initial version of
Wolfram Alpha will address that directly.  It seems that they
do not rely on tagged sources, but I don't know what data sources
they use.  In any case, I think that they have addressed an
important piece of the puzzle, but as Wolfram said, this is
an ongoing project that will never be completely finished.    (022)

AW> Here are a couple of challenge examples for Wolfram Alpha.
 > Both examples are online, where they can be viewed run and
 > edited, using a standard browser.  In each case, end users
 > are able to update the knowledge that is used to answer the
 > questions.    (023)

As I've said many times, I believe that your system of
Executable English addresses an important problem:  writing
business rules for reasoning with and about relational
databases in a readable and executable way.  But Wolfram
is not competing with your system for that business.    (024)

I'm sure that the domain models used in Wolfram Alpha could
access RDBs, and somebody who wanted to use your tools and
their tools together could probably find a way to do so.    (025)

And by the way, I have a much higher respect for Mathematica,
which is a solid achievement, than I have for Stephen Wolfram's
book _A New Kind of Science_, which is a highly speculative
attempt at a theory of everything.  I won't make any predictions
about whether the assumptions at the heart of the NKS book are
true or false, but I'll cite the following discussion:    (026)

    http://www.wired.com/wired/archive/10.06/wolfram_pr.html    (027)

In any case, Mathematica works, and I believe that it is
sufficient to support something like the the claims made
for Wolfram Alpha.  That is independent of the NKS book.    (028)

John    (029)


_________________________________________________________________
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    (030)

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