Upload
cecil-kelly
View
219
Download
1
Embed Size (px)
Citation preview
From Semiotics to Computational Semiotics
State University of Campinas UNICAMP - Brazil
Ricardo Gudwin
Semiotics and Computational Semiotics
Semiotics branch of human sciences, that studies the sciences of signification
and representation, involving mainly the phenomena of cognition and communication on living systems.
Intelligent systems systems that exhibits behavior that can be considered intelligent some of the objectives are the study of the phenomena of
cognition and communication, but now explicitly under the scope of artificial systems
Computational semiotics proposition of a set of methodologies that in some way try to use
the concepts and terminology of semiotics, but composing a framework suitable to be used in the construction of artificial systems, in this case, implementable within computers
Computational Semiotics
Computational semiotics new inborn science, but, there are currently some important contributions that
despite still not complete and definitive, help us in understanding the nature of semiotic processes and allow their synthesis and implementation within computational platforms.
In this work we explore one possible pathway along a set of ideas
evolving around, which proposes one way of gathering the transition between traditional semiotics and computational semiotics, making it possible to synthesize semiotic systems by means of artificial computing devices.
Semiotic Analysis
Semiotics Tool of Analysis - main goal is to understand the semiotic
processing happening in nature Semiotic Beings (interpreters) are already there
living organisms (biosemiotics) human beings (human semiotics)
it is easier to create concepts and apply them to things that do already exist and are already working
Questions should it be possible to use the same conceptual background
in order to synthesize new beings (systems), performing the same semiotic behavior as living/human beings would do ?
What would be the challenges in such an endeavor ?
Semiotic Synthesis
Problem things are not already working so, we should put them to work !
Hidden Problems specify the basic entities involved within semiosis
in a way in which it can be produced within a computer specify the mechanism by which signs are interpreted
there are a lot of intermediary steps that are generally not considered within the context of human semiotics
• How from a scene given by a video camera we discover the objects involved into this scene ?
• Can we talk about signs if the system is still not aware of objects ? Are computational devices able to carry on all sign-processing
that living/human beings are able to perform ?
Semiotic Synthesis
Basic Foundations set up a generic scenario in which semiotic synthesis is going
to be discussed try to get clues on how semiotic processes really happens allow the implementation of a computational version of
semiotic processes Terminology
related to standard semiotic terminology but we don’t want to limit the meaning of terms to human/bio
semiotics Requirement
be careful when applying semiotic analysis to our synthesis scenario
Semiotic SynthesisBasic Foundations
Representation Spaces
INTERPRETER
EXTERNALSPACE
INTERNALSPACE
EXTERNALFOCUS OF
ATTENTION
INTERNAL FOCUSOF ATTENTION
Semiotic SynthesisBasic Foundations
Shareable and Non-shareable spaces
EXTERNALSPACE
(SHAREABLE)
FOCUS OFATTENTION
INTERNALSPACES
(NON-SHAREABLE)
Semiotic SynthesisBasic Foundations
Interpreting Fields
INTERPRETER
EXTERNALSPACE
INTERNALINTERPRETINGFIELD I (x,y,z,t)
(UMWELT)
EXTERNALINTERPRETINGFIELD E (x,y,z,t)
Semiotic SynthesisBasic Foundations
Multiple Internal Spaces and Interpreting Fields
CONCRETESPACE
ABSTRACTSPACE
ABSTRACTSPACE
Semiotic SynthesisBasic Foundations
Interpreting Field concept originated from field theory function (energy function ?) that to each point in space and
time determine a unique value state
External Space interpreting field is continuous (that’s the real world) by definition, is not knowable in its entirety
Internal Spaces accommodate a model of external interpreting field internal interpreting fields are functions that depend on the
type of semiotic synthesis we are trying to model
Semiotic SynthesisBasic Foundations
Sign Everything under the interpreter’s focus of attention (internal
or external) that would cause an interpreter action Interpreter Possible Actions
Change in the focuses of attention (internal and/or external) Determination, for the time t = t+1 of a new value for any
interpreting field (internal or external), at a point (x,y,z) covered by the focus of attention in that space
Interpretant any interpreter action caused by a sign any change in internal and external interpreting fields for time
t = t+1, caused by an interpreter action due to the effect of the sign
External Semiosis
Interpretant of signs happens at the external space
Change in external interpreting field change in environment shareable with other interpreters can act as a sign for the same interpreter or to other
interpreters Happens mainly on interpreters that do not have
internal spaces semiosis in molecules and chemical reactions very simple biological organisms
Can be the final result of a chain of internal semiosis
Internal Semiosis
Interpretant of Signs happens within any of the internal spaces
Signs can be at the external space (semiotic transduction) at the internal space
A typical semiosis chain starts with an external sign generates a set of internal interpretants, that become internal signs generating new internal interpretants, until some of them become an internal sign that generates an external interpretant
Information, Signs and Knowledge
Signals and Information signals - values of parts of interpreting fields that can be
differentiated (distinguished) from other values information - meaning of signals
Example suppose that E (x,y,z,t) has a counter-domain like [0,5] but, due to sensor limitations, the interpreter is only able to sense
values in {0,1,2,3,4,5} then, values like 2.3 or 2.2 would equally be understand like 2 so, the information that those signals convey is tied to only 6
discrete values
Signals only describe states they do not cause any actions
Information, Signs and Knowledge
Once signals are able to cause actions they become signs
The information they carry associated to the actions they cause is then called knowledge
Signals - Information Signs - Knowledge Region under a focus of attention of some space
sign knowledge unit
Things to Remember
External Interpreting Field is infinite, continuous and probably take values on continuous
sets can not be known as a whole can be known in parts, with approximations
The only way we are able to know the external interpreting field is due to sensors
The most basic knowledge units that can be stored into internal interpreting field is of sensorial type
Internal Interpreting Field (Concrete Space) our best model of external interpreting field
Things to Think About
Storing sensorial information is not efficient We need better models
Basic mechanism The notion of “Entity”
From sensorial knowledge units the system must try to represent the same external interpreting
field as a collection of entities Entities
may have attributes that would change in time Occurrences
model the change in entities attributes Sensorial knowledge, entities and occurrences
grouped to represent situations
A Hierarchy for Knowledge Units
Indexical
Iconic
Symbolic
Symbolic
Iconic
Knowledge Unit
ArgumentativeDicentRhematic
GenericSpecific
Entities
GenericSpecific
Occurrence
Generic
Specific
Sensation
Analytic Synthetic
AbductiveDeductive
Inductive
Simplification of the Model
Instead working in general spaces and interpreting fields restrict ourselves to memories and places assign sign processing to micro-interpreters
KNOWLEDGE UNITS
SPACE
PLACE
Micro-Interpreter
Micro-Interpreter’s Responsibilities choose the knowledge units that will be used (focus of
attention) eventually destroy them after use create new knowledge units using information contained in
earlier ones
Interpreter
Knowledge Units
Building Intelligent Systems
Using multiple micro-interpreters in cooperation to
each other processing
knowledge units
Conclusions
Dyadic or Triadic ? Some people would say that we are building a dyadic model for a
sign But, there is some kind of “mediation”, due to:
the focus of attention mechanism the influence that some knowledge units may have over the processing
of other knowledge units (catalytic knowledge units) We still need to do further reflections in order to have a better
position on this issue This is not the final word regarding Computational Semiotics
it is only a first exercise in order to get insights to the problem of semiotic synthesis
computational implementations of such model indicate that, up to some point, it is worth the value of working on it