| |
|
Department Colloquium: Osamu Fujimura |
-
- Osamu Fujimura. Wednesday, 17 January
2007. 4-5pm. 618 Crawford Hall. Title: "Symbolic representation
in language and biology." Dr. Fujimura is Professor
Emeritus, Ohio State University. He established and directed the
Research Institute of Logopedics and Phoniatrics at the University
of Tokyo. In 1965, he led the speech section of the Acoustical Society
of Japan, receiving its Award for Distinguished Service in 1999.
He collaborated for four years with the research laboratory in electronics
at MIT and for two years with the Royal Institute of Technology
in Stockholm. For was a director at Bell Laboratories (AT&T)
for fifteen years, serving in the Departments of of Linguistics
and Speech Analysis and Artificial Intelligence Research. He was
professor in the Department of Speech and Hearing Science, The Ohio
State University for fifteen years. He is currently preparing a
new introductory book on speech science.
Abstract:
The emergence of intelligence, specifically in the case of homo
sapiens in relation to the development of language, may be assumed
to be inherently related to the innate metagrammatical system of
human infants as a special biological framework. This initial condition
of human cognitive faculty was suggested originally by Noam Chomsky
to explain the specially efficient development of the linguistic
competence in human children. On the other hand, the evolution of
biological species is traditionally ascribed to formation of successful
genes as particular linear strings of molecular elements along with
other accidentally produced variants in competition, combined with
a subsequent selection process accepting more suitable forms and
eliminating less suitable forms for survival within the environment
that can provide limited resources. Can the speed of evolution,
particularly in the recent phase from higher primates to homo sapiens,
be explained by randomly produced variants combined with this binary
choice between survival or extinction of the embodiments of the
genes through generations? Should the concept of evolution be limited
to the change from generation to generation via reproduction? Should
the process of selection and survival be limited to competition
for life of individuals eliminating existing copies that are already
embodied?
Can there be a principle that evaluates the merit of a promising
gene among variants and facilitates its embodiment before they are
materialized? Is there a principle that filters or restricts the
generation of genes, or their specifications encoded as genomes,
in terms of some level of abstract representation that serves as
genomic codes and for their possible changes? Can there be some
criteria for selecting good qualifications (in terms of features,
as in linguistic theory) of variants in terms of their representations
(particular organizations of elements), in such a way that once
such selections of features are made, their combinations may develop
effective changes explosively rapidly, as we observe in the development
of human knowledge? Provided appropriate basic hardware design that
leads to such selective combinations by proper choices in the form
of organization of features as software, we may expect, as in the
case of human-made computation machines, extremely rapid “progress”
toward a direction that seems to be “purposeful” through
merely random processes. For such changes to take place, and for
effective criteria to be applied for “useful” selections,
physical (chemical) processes that are based on interactions between
contiguous elements in the genome may not be expected to be sufficient.
Global principles of interaction, based on feature matching, may
be more effectively applicable. We remind ourselves that such global
feature interactions are characteristic of syntactic and phonological
organization principles in theoretical linguistics. If such principles
operate based on innate biological properties of humans, where do
they come from? Are there basic principles that govern, perhaps
just as a matter of accidental selection or according to some deeply
seated principle of our universe, the organization of biological
systems and emergence of life, as well as the linguistic systems,
in relation to the development of the nerves and their physiological
network organization?
The physical structures of genes are being identified using advanced
technology, and new insights are being obtained discovering some
parts and properties within the genome to have specific functions
for its embodiment as proteins. We wonder if the organization principle
of genomes may be described by an abstract representation of some
structure, perhaps in some sense similar to the phrase structure
of sentences in syntax, that underlie particular linear strings
of base elements but contains more crucial information. The organizations
of biological organs may be associated with a hierarchical structure
of proteins, building the more complex out of the more simple, propagating
crucial features. Such an abstract structure may be effective in
producing possible (i. e., grammatically acceptable and biologically
survivable) preselected variants produced by random perturbation.
The structure may automatically prevent, under normal circumstances,
changes that cannot (or should not, in some sense) occur, and such
constraints may not be apparent in the local properties of the genomic
codes; for example, most random changes deviating from the legitimate
patterns may result in blocking its implementation, except in pathological
developments that are likely to be removed by Darwinian selection.
In this connection also, we should be able to understand why the
actual lives are mostly normal.
According to the current (Chomskyan) syntactic theory, there are
simple principles underlying the design of human language in general,
and the diversity of natural languages as we observe are the results
of selecting one of a few discrete values of each of a small number
of parameters for particular languages. Human language acquisition,
according to the surrounding spoken language for the newborn baby,
works as a process of setting such parameter values of the innate
system of language based on the grammatical principles by the use
of available very limited empirical data. The representation of
the grammatical structure, however, uses a complex hierarchical
organization of each sentence rather than a linear string of constituent
elements directly observable in speech signals of the sentences
that are heard.
This abstractness of representation with the underlying hierarchical
organization of functional units characterizes human language. Do
we have corresponding principles and selection of parameter values
within any abstract framework for specifying genes in the process
of genetic evolution? It would not be particularly surprising if
nature employs common principles or elemental operations for the
design of genetic codes on the one hand and the linguistic cognitive
system on the other. For example, does recursion as a general computational
principle characterize the structural representation of genomes
or their manifestations such as proteins and chromosomes, like in
syntactic specifications of sentences? Does the endocentricity (self
embedding based on designated critical features) in the process
of syntactic projection play a central role also in genomic code
formation processes? There must be a way to represent genomic structures
by a general biological principle, with parameters to be set for
individual species.
|
