Another fantastic citation worth it's weight in gold and definitely
relevant to the long-term goal here of creating an algorithmic means
to express - and then operate on - biomedical knowledge! Many
thanks, Bob. I've already passed on your "hedging" reference to
several other colleagues.
Having said that, I do *strongly* feel its very important to make a
distinction between linguistic analysis and ontology development.
The two are very different animals - albeit with intricately
overlapping interests in the nature of how one expresses semantic
information for both humans and machines. To be to hasty to cast
these resources all as woefully imperfect, is to throw out the baby
with the bath water - when it comes to producing accepted practices
and tools to support machine processing of semantic content.
The problem as I see it if you convolve the two is they come with
significantly different a priori assumptions and expectations for
their use. When one looks at an analysis of the lexicon used in a
specific research article to describe a specific statement - the
equivalent of which can be expressed as an RDF triplet - one needs to
know the implied, allowable PROCESSING (as Xiaoshu has been
stressing) for such a triplet is very different than when one has a
triplet expression derived from a "universal" relationship
represented in an ontology.
This is the critical point I've been trying to make throughout this
thread - and other related threads on the BIOONT task force.
This distinction may be less evident and/or relevant when you are
just trying to communicate this information to another human being,
but when the goal is a provide it in a useful way to a machine
algorithm that is expected to compute on it in some useful way, the
distinction is critical.
When folks are first coming to understand the nature of the various
efforts that have been underway for at least 60 years to come up with
more formal means to represent the SHARED concepts biomedical
scientists intend to evoke when they are communicating their findings
- and theories - with others - it can be a helpful aid to cast these
lexical resources and complex ontologies as lying along a continuum
of semantically-oriented resources. As Bob is implying in this post,
they lie much closer together than most would like to think and as a
collective don't do a particularly good job at deterministically
expressing the subtle, fine-granularity of meaning either embedded in
or implied by specific scientific communications. They also both
bring with them many problematic tasks, if the goal is to keep the
terminology or ontology consistent with the "bleeding edge" of
knowledge in every micro-domain of a scientific field. These facts I
most definitely agree with.
I would say there is quite a bit of semantically-oriented data
management & knowledge mining "low hanging fruit" available, IF one
keeps the distinction between these two types of knowledge resource
separate.
Again - lexical/linguistic resources are constructed in very
different ways and bring with them a very different set of a priori
assumptions and ultimate goals than the development of formally, well
founded ontological frameworks cast in the Leibnitzian realm of
providing a means to compute on "meaning". I would also stress that
compute on meaning does not only mean to interpret and re-combine
complex logical assertions (1st or 2nd order). These tools can also
be used to great effect to look for gaps and inconsistencies in the
semantic graph you are constructing - the bread & butter maintenance
tasks groups like the Gene Ontology Consortium need to automate as
much as possible. This is certainly true for the work we are doing
in on the BIRN Ontology Task Force.
I would also add that it is usually the case when mapping out new
domains to include in an ontology, you typicall start with an
analysis of the lexicon in that domain - by collecting and analyzing
the terms used by scientists in that field - down to deep levels of
granularity. You then go about the task of organizing the lexical
relations, as a means to come up with a more complete and consistent
representation of this lexicon - and all of the lexical variants and
inter-relations (e.g., synonyms, homographic homonyms [a particularly
nasty beast for algorithms to disambiguate], meronym-holonym pairs,
hyponym-hypernym pairs, etc.). if one goes about this task informed
by the extensive work from the fields of linguistics and psychology
of language (think - WordNet - as just one example), then the task of
using this lexical framework as an outline for an ontology is made
ever more accessible. In this context, analyzing the biomedical
lexicon and moving toward ontological frameworks - I would say way
too little attention has been paid to the issues Bob has been
referring us to. I think a great deal could be gained from
incorporating an understanding of these complex issues - hedging and
the limits of knowledge representation - into the process.
I do think, however, that should following harvesting the "low
hanging fruit".
Just my $0.02.
Cheers,
Bill
On Jun 19, 2006, at 5:11 PM, Bob Futrelle wrote:
I would suggest that both natural language *and* ontologies are views
of (possibly shallow) underlying knowledge. This knowledge is
difficult to characterize. It is also difficult to achieve agreement
on it within or across communities.
I find the following study sobering. Don't be misled by the term
"folk". Today's science is tomorrow's folk science.
- Bob Futrelle
----------------------------------------------------------------------
-
Abstract
Cognitive Science: A Multidisciplinary Journal
2002, Vol. 26, No. 5, Pages 521-562
(doi:10.1207/s15516709cog2605_1)
The misunderstood limits of folk science: an illusion of
explanatory depth
Leonid Rozenblit - Department of Psychology, Yale University
Frank Keil - Department of Psychology, Yale University
People feel they understand complex phenomena with far greater
precision, coherence, and depth than they really do; they are subject
to an illusion—an illusion of explanatory depth. The illusion is far
stronger for explanatory knowledge than many other kinds of knowledge,
such as that for facts, procedures or narratives. The illusion for
explanatory knowledge is most robust where the environment supports
real-time explanations with visible mechanisms. We demonstrate the
illusion of depth with explanatory knowledge in Studies 1–6. Then we
show differences in overconfidence about knowledge across different
knowledge domains in Studies 7–10. Finally, we explore the
mechanisms
behind the initial confidence and behind overconfidence in Studies 11
and 12, and discuss the implications of our findings for the roles of
intuitive theories in concepts and cognition. (c) 2002 Leonid
Rozenblit. Published by Cognitive Science Society, Inc. All rights
reserved.
Bill Bug
Senior Analyst/Ontological Engineer
Laboratory for Bioimaging & Anatomical Informatics
www.neuroterrain.org
Department of Neurobiology & Anatomy
Drexel University College of Medicine
2900 Queen Lane
Philadelphia, PA 19129
215 991 8430 (ph)
610 457 0443 (mobile)
215 843 9367 (fax)
Please Note: I now have a new email - [EMAIL PROTECTED]
This email and any accompanying attachments are confidential.
This information is intended solely for the use of the individual
to whom it is addressed. Any review, disclosure, copying,
distribution, or use of this email communication by others is strictly
prohibited. If you are not the intended recipient please notify us
immediately by returning this message to the sender and delete
all copies. Thank you for your cooperation.