previously Copyright 2007 David Dodds

My first published material in the area of technical metaphor was in
1981,a paper at the 1981 Annual Meeting of The American Association
for the Advancement of Science. AAAS This is the organization which
publishes the prestigious journal 'Science'.

In the next posting or the one after that I will post the content of
that paper. It discusses what technical metaphor is via some examples.

In this posting we see the content=

slides from the International AI 2006 computer conference
http://ww1.ucmss.com/books/LFS/CSREA2006/ICA8076.pdf
where I presented the topic below:

David Dodds
Ontologies and Conceptual Metaphor in Autonomous Robotics

David Dodds
TOPICS:
Ontologies
Conceptual Metaphor
Autonomous Robotics

Autonomous Robotics
A robot system such as OPRS permits one to set up
pre-programmed "notions" (representations in memory)
for paths, way-points, obstacles, for example and derive
the presence of these by examining sensor data. OPRS
usually is run in reactive mode, but it is capable of a
meta-level mode where deliberative processing can occur.
Usually way-points and path usage are processed
reactively at initialization time.

Autonomous Robotics 2
Planning the path of travel, a typical job of on board
computing systems in an autonomous mobile robot, might
be performed using the A* approach. A metaphorical planning
system can choose to use higherlevel geo-spatial objects
in addition to points, lines / paths used in standard A*.

Autonomous Robotics 3
• Plans which substitute inexact values and inexact actions
for precise ones give the plans a flexibility closer to that
of biological systems, who never know precisely where
their limbs are nor distances from themselves to things
of interest in the world. This <it> situated-ness </it>
(relating both perceptions and measurements of the world
and reasoning relationships to oneself or substitute-other
perceptual origin) is what provides biological systems with
such reasoning power to cope with a complex world.

Autonomous Robotics 4
• Context is which features and dimensions of all those
which the system "knows" of are relevant, salient to
the moment or task at hand.
• Similarly when we were babies thrashing about in the
crib we noticed the correlation between our reach
(general arm movements for that matter) and what
we saw with our eyes. That conceptual correlation
provided us with a sense of space, of depth (3D
volume, extent).

Autonomous Robotics 5
• Measuring the distance from the laser range finder
on a robot to the current location of the robot's hand
does not provide a sense of depth.
• Many such measurements would have to be taken
(with the hand at different locations obviously) and
these would have to be combined with concomitant (robot
limb) proprioceptive information (perhaps angles at
inflection points and motor information as well).
• Metaphor-based planning is potentially much more
flexible than exact-match approaches

Autonomous Robotics 6
•Metaphors are used to "explain" or model phenomena
or even concepts. They are productive because they
are inexact, imprecise and yet capture or <IT> represent
the salient aspects or features </IT> of the phenomena
or concepts.
•High-quality computer games use a lot of code to
"process physics", like gravity, gas expansion, etc. and
generate reasonable action / behaviour to make the games
more like the real physical world. Autonomous mobile
robots benefit from the same processing capabilities as
it helps to interpret sensor data and generate
expectancies for what may happen.
•Metaphors of various types help reasoning by providing
a known description or process to address reasoning about
something which is not a known description or process.

Autonomous Robotics 7
• Words are containers for thoughts, which themselves
are objects.
• It is the ability to functionally process inexact things
which allows this very useful abstraction, the substitution
of something <IT> similar </IT> or <IT> corresponding </IT>
for unknowns. It allows reasoning about things that are not
perfectly familiar by using those that are.
• Neither modeling nor abstracting (metaphors) can be
performed willy-nilly but must be done in a principled manner
in order that the result be functional.

Autonomous Robotics 8
• Part of the solution is to use relevancy / conditionals,
to use relevant domains and dimensions / aspects. Partial
ordering of the innards of the metaphor is the answer to
this.
• The sciences of astronomy and physics have long used
conceptual metaphor productively, in such terms as "event
horizon", "gravity well", "curved space", "time dilation".
• Productive conceptual metaphors can be used functionally
on computers. This may be done by employing ontologies
which provide requisite background knowledge to the
system and qualitative logic processing.

Autonomous Robotics 9
• In the case of "The squirrel ate the acorn near the foot
of the tree.", we have fuzzy and metaphorical elements in
the statement.
• The basis for the context which provides a scale for "near"
in this case is "tree", specifically the comparison of the general
distance from the (actual or estimated) center of the tree
trunk to the (general or average, or even believed to be) outside
edge of that tree, with some actual (or estimated) distance of
the acorn from the (edge of) the tree.
• Foot of tree, foot, is a name which can be used in ontologies as
a linguistically derived term.

Autonomous Robotics 10
•When we look at the CYC knowledge-base we see that it has
defined a number of valuable useful knowledge items to do
with foot.
•CYC may be used to infer tacit information, that is information
which is not explicitly present in an input / data set.
•This ontological semantic knowledge is valuable for it allows
the program we are discussing in this paper to identify / locate
a foot in a visual data set yet without being committed to a
particular photograph of one, nor drawing or X3D three-dimensional
data set of a foot.

Autonomous Robotics 11
• <IT> relevant transfer </IT> of both goals and actions from
the <IT> activity </IT> of an "easter egg hunt" to that (activity)
of locating and acquiring dispersed objects of value at unknown
and possibly hidden locations.
• Another metaphorical plan "Don't get too close to a deep hole."
• Looking at the comments in the CYC ontological items you may
see that the natural language terms for fuzzy predicates, such
as ABOVE, BELOW, Directly, Generally, Up, Higher occur there.
Computing With Words technology (Zadeh et al.)
can provide real computations employing these concepts / predicates,
as illustrated above with the fuzzy term "IsNear". ((The next
posting contains the equation for IsNear, so
that it may actually be calculated in a robotic system
[perhaps one using OPRS as mentioned
earlier] using actual data from its sensor system.
You will then be able to see how
"too close" and "deep hole" are actually computed.
i.e. - strings given procedural "meaning"))

Autonomous Robotics 12
• Metaphors like "foot of.." use the same computing
principles as these. By detecting and applying relevant
contexts, such as which dimensions make sense to
compare and process, these computations can bridge
fuzzy systems / Computing With Words with
ontological technologies. These contexts provide a
measure / identity of the "partial"
aspects of "partially structured metaphors."
• Fuzzy actions, when we place things "by eye" we
place them "generally near enough"
to the (perceived) actual centroid of the receiving object,
we never measure exactly where the centroid of things
is to place objects.

Autonomous Robotics 13
•SVG programs can be parsed with an XML parser and
so semantic-web type technologies may examine SVG
programs directly and discover semantic knowledge
therein. ((Such as relative spatial positions.))
Through such parsing and discovery the actual SVG
program can be translated into fuzzy representations
so as to allow them to be processed via
Computing With Words technologies.
•(Gravity Well) A hypothesis forming / "discovery"
system could take the data and generate theories.
•Years ago (1986) "Boyle's Law" was rediscovered
by a discovery program
(BACON) developed by Pat Langley.

In an upcoming posting the code of an SVG program
which produces a bar-chart will be shown.
If you have an SVG competent viewer (not IE, unless
you use the Adobe plugin which they have
discontinued) you will be able to view / run the shown SVG
code to produce a visual output. Also shown in these
postings will be parts of ontologies of
Space, Time, Material and several others, from
NASA JPL SWEET, and CYCorp.
It will be clearly shown how ontologies combined with logic
(via programs) can "interpret" SVG pictures. An SVG
bar-chart (visual and source code), and embedded semantic knowledge
tie together these things. Stay tuned.
PHP5 can be used to parse XML documents and that means
SVG programs as well. A future
posting will discuss PHP on Rails, a PHP framework which uses
the Rails technology.
Everyone knows about Ruby on Rails, right?.
and yes, PHP fans, there will be discussion of
PHPEclipse. Also coming down the road is a discussion of
dynamically-graphed hyperthreaded
programs (ie dataflow) in such systems as 64 Studio, using tools like
JACK server and (cornucopia of) friends.