Adrian, Ferenc, Ron, and Jack, (01)
Before commenting on your notes, I'd like to mention that I came
across a paper co-authored by David Ferrucci in 2008, which was
shortly after they began the Watson project: (02)
http://arxiv.org/ftp/arxiv/papers/0809/0809.0680.pdf
The Prolog Interface to the Unstructured Information
Management Architecture (UIMA) (03)
Following is the concluding paragraph of that paper: (04)
> The UIMA generic Prolog annotator allowed us to develop faster and easier
>pattern
> matching rules for natural language analysis in a language familiar to our
>developers
> and users (i.e., the Prolog language), the Prolog engine being transparent to
>the UIMA
> pipeline (i.e., completely integrated in the pipeline), while having access
>to state-ofthe-
> art semantics and proving effective on question analysis (i.e., time and
>results).
> We implemented interfaces for various rule systems: the UIMA-Sicstus Prolog
> interface (using the PrologBeans library) [6], the UIMA-SWI Prolog interface
>(using
> the JPL library) [7] and the UIMA-InterProlog translator for used by XSB [8]
>and
> Yap Prolog [9] systems (using the Interprolog library [10]). Our applications
>of this
> annotator include: complex rules for question analysis, shallow semantic
>parsing, and
> tools for development and testing UIMA analytics. (05)
The paper is only five pages long, but it gives a bit more detail about
the kinds of things that Watson is doing. And I am very happy to see
that they use Prolog, which is an outstanding language to use for this
kind of application. (06)
In fact, Prolog is the primary language that we use at VivoMind, because
it is highly flexible and can be quickly adapted to either informal
processing (along the lines used by Watson) or precision analysis
(as needed for formal logic). We also use C, but only for heavily
used, well tested algorithms that can be frozen in low-level code. (07)
AW
> Watson is of course a major achievement, as it demonstrated by
> comfortably winning Jeopardy. It's now official that IBM sees
> Watson has having potential in Medicine -- it could read the
> biological-medical literature and outperform the Doctors.
>
> However, for these purposes, there's a key difference between
> Jeopardy and Medicine. In Jeopardy, humans know the right answers
> -- the city of Toronto is not in the USA.
>
> In Medicine, humans don't have consensus answers to new questions
> (e.g. What is the best treatment for multiple sclerosis). So,
> whatever algorithms Watson uses will lead to new medical knowledge
> that humans cannot easily check by thought experiments. (08)
That may be true. But there is a huge amount of knowledge in the
medical literature that a practicing physician can't possibly know.
Even a research physician can only keep up with the literature in
his or her own specialty. There is no single MD in the world who
can know all or even most of the consensus answers. (09)
I would not expect Watson or any other computer system to produce
a definitive answer to any medical question. But what it could do
very well is produce several alternatives with its own confidence
ratings for each *and* with pointers to the literature for the
physician to verify. That would be immensely valuable. (010)
FK
> Surely, if the knowledge base and learning algorithm of Watson is
> based on a dynamic, but single algorithm as opposed to a data base
> ... Call the undertaker (011)
First of all, Watson has a very wide range of different algorithms.
But in any case, neither Watson nor any other computer system being
designed today would ever replace a physician. Its primary role is
to serve as a super search engine to find relevant knowledge that a
physician might not be aware of. The final decision about treatment
is always the responsibility of the physician. If anything goes
wrong, the human MD is sued, not Watson or IBM. (I worked at IBM,
and I know that IBM management is highly allergic to law suits.) (012)
RW
> It appears that Watson can give you a lot of insight into how it
> arrived at an answer including the various parallel processes that
> were done. It probably can do a much better job of this than most
> humans since we quickly forget the bad ideas and fruitless paths
> whereas Watson remembers them all. (013)
Yes. Think of Watson as a super Google that keeps track of everything
and evaluates the alternatives. But the human MD makes the decision. (014)
JP
> ... just the task of thinking through *how* to organize resources
> for Watson to deal with them is, itself, an important learning
> opportunity. (015)
I agree. (016)
RW
> It will be interesting to see how "ontologists" make the shift from
> being "owltogists" to "Watson feeders". (017)
Watson is much more flexible than OWL. A knowledge engineer working
with OWL is forced to state every point very precisely in an exactly
*decidable* way. But most of the knowledge in every field is vague,
flexible, and rarely, if ever, *decidable*. (018)
There are very specialized domains (microtheories) for which OWL and
other formal logics are valuable. But the overwhelming amount of
knowledge in the world is *unstructured* -- the first letter of UIMA. (019)
I believe that the combination of Prolog with UIMA (or something
like it) is much better suited to processing the vast resources
of the Web than OWL. (020)
John (021)
_________________________________________________________________
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 (022)
|