ontolog-forum
[Top] [All Lists]

Re: [ontolog-forum] IBM Watson's Final Jeopardy error "explanation"

To: ontolog-forum@xxxxxxxxxxxxxxxx
From: "John F. Sowa" <sowa@xxxxxxxxxxx>
Date: Thu, 17 Feb 2011 21:45:08 -0500
Message-id: <4D5DDD34.4040108@xxxxxxxxxxx>
Patrick,    (01)

I really don't understand that complaint:    (02)

PD
> ... what you were trying to do in your first line is to put the
> naming issue firmly in the hand of the AI community.
>
> Sorry, too old a trick to allow to pass without comment.    (03)

Following is the opening of my previous note, in which I responded
to Peter B's comment that "artificial intelligence is neither."    (04)

JFS
> That's a quibble about a name.  Many people, including me, have stated
> such quibbles from time to time, but they're irrelevant.  They're as
> pointless as the behaviorists who objected to the name 'psychology'
> because it implies an unobservable psyche.    (05)

Notice that I explicitly dismissed Peter's complaint as irrelevant.
That's too obvious to be a trick.    (06)

PD
> Looking at "accomplishments," the fine folks at IBM designed a black
> box system (to outsiders) that answers questions on a game show.
>
> Gee, why am I not impressed?    (07)

Actually, there is sufficient documentation available that anybody
knowledgeable about AI and NLP who had access to a team of 25 people,
a supercomputer with 2820 cores, four years time, and a few million
dollars to pay salaries could reproduce something as good or better.    (08)

PD
> Had the team published its NLP processing algorithms, which were no
> doubt innovative, that would raise all boats.    (09)

I'm sure there was some innovation in algorithms, but nothing that
talented AI experts couldn't invent just as easily as they could copy.    (010)

For anybody who would like to reproduce their approach, I suggest
starting by reading David Ferrucci's slides from 2006 about UIMA
(the Unstructured Information Management Architecture):    (011)

http://ontolog.cim3.net/file/resource/presentation/DavidFerrucci_20060511/UIMA-SemanticWeb--DavidFerrucci_20060511.pdf    (012)

Following is the talk that explains the slides:    (013)

http://ontolog.cim3.net/file/resource/presentation/DavidFerrucci_20060511/UIMA-SemanticWeb--DavidFerrucci_20060511_Recording-2914992-460237.mp3    (014)

For more about UIMA, which IBM released to open source under the care
of the Apache Foundation, type "UIMA" to Google.    (015)

And following is Ferrucci's recent talk about the DeepQA project for
building and extending that foundation for Jeopardy:    (016)

http://www-943.ibm.com/innovation/us/watson/watson-for-a-smarter-planet/building-a-jeopardy-champion/how-watson-works.html    (017)



For further documentation, see the following technical report
by Ferrucci et al. while they were two years into the project:    (018)

http://domino.watson.ibm.com/library/cyberdig.nsf/papers/D12791EAA13BB952852575A1004A055C/$File/rc24789.pdf    (019)

At the end of this note is an excerpt from a note I sent to
Ontolog Forum yesterday with further discussion about Watson,
based on an article about it published in AI Magazine.    (020)

Given the above information, what I summarize below, and the
money and hardware available to Ferrucci's team, quite a few
good AI groups around the world could build a system that would
perform as well or better than Watson.    (021)

John    (022)

-------- Original Message --------
Subject: Re: [ontolog-forum] IBM Watson & source code
Date: Wed, 16 Feb 2011 11:29:38 -0500
From: John F. Sowa    (023)

MB
 > From what little I've seen it does rather seem to be a triumph
 > of statistics over semantics.    (024)

Statistics certainly plays a big role, but the relative role
depends on what you mean by 'semantics'.  I would say that Watson
represents a triumph of 1980s style of NLP, and I'd summarize
the influences in one line:    (025)

    Michael McCord + Roger Schank + statistics + supercomputer    (026)

For McCord's influence, I'll quote the following passage from the
article in AI Magazine:    (027)

Page 11, http://www.stanford.edu/class/cs124/AIMagzine-DeepQA.pdf    (028)

 > The DeepQA approach encourages a mixture of experts at this
 > stage, and in the Watson system we produce shallow parses,
 > deep parses (McCord 1990), logical forms, semantic role labels,
 > coreference, relations, named entities, and so on, as well as
 > specific kinds of analysis for question answering.    (029)

In the 1980s, McCord had written an excellent parser in Prolog
with a good grammar of English.  In the 1990s, he rewrote the
parser in C for better performance.  Apparently, that's the
"deep" parser they use for Watson.    (030)

Also in the 1980s, Roger Schank made the claim that axioms in
logic are irrelevant, but large volumes of background knowledge
are essential for language understanding.  He also said that
machine learning is essential for NLP, since every text says
something new that must be added to the background knowledge.
Unfortunately, the technology at the time was too slow, and
the available resources of machine-readable texts were woefully
inadequate to support Schank's claims.    (031)

Watson also uses ideas developed in the 1990s and later, but
one could say that the Watson-style of semantics is a high-speed
implementation of a Schankian style of NLP.  The crucial technology
that Schank did not have is a supercomputer plus huge volumes of
preprocessed material indexed and accessible via a relational DB.    (032)

I certainly won't downplay the importance of statistics, which
are essential for many aspects of Watson:  evaluation of what
is relevant, estimating the confidence in an answer, and most
especially for techniques of  machine learning.  But learning
is also one of the aspects that Schank emphasized years ago.    (033)

In short, Roger Schank's emphasis on informal methods of
processing and using large volumes of background knowledge,
case-based reasoning, and machine learning are much closer
to what Watson is doing than any logic-based method --
either Richard Montague's formal logic for NLP or Lenat's
enormous formal ontology for Cyc.    (034)

Yet Watson does use some logic, however.  It's just not the
main focus.  Statistics and heuristics are more important.    (035)

KJ
 > As Watson turns out to be very successful this leads again to the
 > question which role ontological categories play...    (036)

Watson does use WordNet and other lexical resources.  That is
important for selectional constraints on permissible combinations,
but Wordnet and similar resources have very few axioms.    (037)

_________________________________________________________________
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    (038)

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