« Return to Thread: Wolfram Alpha

Re: Wolfram Alpha

by Robert Spillers :: Rate this Message:

Reply to Author | View in Thread

Some parts of this message have been removed. Learn more about Nabble's security policy.
John,
You may remember that this is the sort of approach IBM  (Bryan Sivak, Ed Cooper, Martin van den Berg, Andras Kornai and others) used in the mid 1990's.  We (I was the Program Director) worked with Doug Lenat and Cyc, you were a consultant on the project as was John McCarthy. I commissioned a verb lexicon from CSLI at Stanford using WordNet as an ontology - we intended to commission an ontology from Cyc.  After I retired I persuaded IBM to release the lexicon to Stanford to be put in the public domain. I understand that a number of start ups have used the lexicon in their products (including VivoMind).  

IBM failed to see the value of the project and canceled it.  I had worked for IBM for thirty years, thought that was enough and retired.  All of the people who worked on the project eventually left IBM. 

Bryan and Ed had worked on a project at the University of Chicago involving the four  wh.. words.  I hired them when I was on campus to attend my godson's (and Bryan's) graduation.  Bryan and Ed founded InQuira.  Andras has been Chief Scientist at several companies.  Martin along with Livia Polanyi were original employees at PowerSet which was recently bought by Microsoft.

Bob


From: John F. Sowa <sowa@...>
To: cg@...
Sent: Tuesday, March 10, 2009 10:34:48 PM
Subject: Re: [CG:] Wolfram Alpha

John V. and Frithjof,

JV> Looks like we can all pack up and go home.

FD> it is somewhat unsatisfying to not having more information
> about the technologies behind ...

Doug Lenat has a good overview of their system:

  http://blog.cyc.com/2009/03/wolfram-alpha.html

From that note and other sources, I would assume the following
summary of their technology:

1. They are using databases of information that they have put
    together from various sources (including some they access
    dynamically for data about the weather or the stock market).

2. They use a large collection of patterns of phrases (i.e.,
    templates) for recognizing four question types -- who, what,
    when, and where (with or without those actual question words).

3. They have built mathematical/logical models (i.e., ontology
    plus a reasoning or computational method) for using the
    resources in #1 to answer questions of the types in #2.

4. Their computational and logical methods are based on the
    very large and very sophisticated resources of Mathematica,
    which is the primary product of the Wolfram company:

    http://www.wolfram.com/

The original Mathematica product, which they began selling in 1988,
was based on Prolog.  Since then, they have developed their own
logic programming system, which has added many functions that go
far beyond plain vanilla Prolog.

The Mathematica product is widely used by major scientific and
engineering R & D institutions around the world.  (By the way,
we use it at VivoMind for testing our mathematical algorithms
before translating them to a lower-level language for better
performance.)

John


---------------------------------------------------------------------
To unsubscribe, e-mail: cg-unsubscribe@...
For additional commands, e-mail: cg-help@...

 « Return to Thread: Wolfram Alpha