|
OBJECTIVES
Objectives of the
research
a.
Design,
implementation and testing of a neural network model for the acquisition of
grounded sym and symbolic relationships. The model will demonstrate how it
is possible to manipulate the neural network architecture and the learning
task to simulate the acquisition of symbolic knowledge and language-like
symbols.
b.
Delivery
of various simulation models. Some use evolutionary neural networks and
evolutionary robotics for studying the emergence and evolution of syntax in
multi-agent systems. These models will have parameters to test the
conditions that favour the evolution of syntax, and to test the
co-evolution of language and brain complexity. A different approach is used
in a further multi-agent model. This uses a parser of (simple) English
language to study the emergence of language universals (e.g. word order)
under cognitive constraints (such as memory).
c.
Execution
of extensive computational simulations with models as specified in
objective a and b. theoretical hypotheses of symbols acquisition, syntax
evolution, and language-brain coevolution will also be investigated. The
analysis of the simulation results will be published in various conference
and journal papers. Some publications will also address the practical applications
of these models in Robotics and multi-agent systems.
WORKPLAN
Phase I: Neural
network models for the acquisition and transfer of grounded symbols
The beginning of the research was dedicated to
the review of existing models and approaches in language evolution
modelling. This also corresponded with the PI’s work on the book editing,
and the PhD student (Huck Turner) preparation of the review paper (appeared
in the same book). In addition, both staff worked on the design of the
simulation program for the language universal model (Turner) and the
verb-noun model (PI).
Deliverables:
-
1
edited book (Simulating the Evolution of Language)
-
2
book chapters (Turner’s review; Cangelosi & Turner in Borghi-Iachini)
-
1 journal
paper (IEEE Trans. Evolutionary Computation); 1 conference paper (Evol.
Language 2000)
-
Simulation
software (visit Software
download page)
Phase II: Evolutionary
models for the origin of syntax
During the second stage, the PI conducted a
series of experiments on the evolution of nouns and verbs (in collaboration
with Domenico Parisi) and the Baldwin effects simulations (with Steve
Munroe). The PI was also directly involved in the simulation on symbol
grounding through the on-line collaboration with Thomas Riga and Alberto
Greco in Genoa. Huck Turner completed the implementation and testing of the
first part of the model (parser).
Deliverables:, 1 journal abstract (Brain and
Cognition); 2 conference papers (International CogSci Meeting, Italian
CogSci conf.); 1 journal commentary (Behav. Brain Sciences); 4 conference
abstracts (2 Language Evolution 2002, Brain Culture Conf., CogSystems
workshop).
Deliverables:
-
2
journal papers (Artificial Life, Italian Journal of Linguistics)
-
1
journal abstract (Brain and Cognition)
-
2
conference papers (International CogSci Meeting, Italian CogSci conf.)
-
1
journal commentary (Behav. Brain Sciences)
-
4
conference abstracts (2 Language Evolution 2002, Brain Culture Conf.,
CogSystems workshop).
Phase III: More on
syntax evolution and the co-evolution of language and brain
Simulations on: Synthetic brain imaging of the
processing of verbs and nouns (with D. Parisi), evolutionary robotic model
(with D. Marocco), word-order and memory constrains (with H. Turner).
Deliverables:
-
2
journal papers (Brain and Language, Phil. Trans. Roy. Soc.)
-
2
conference papers (Bio-robotics workshop, ICANN-ICONIP03 conf.)
-
1
book chapter (in Usai & Zanobini)
-
6
conference abstracts (Lang. Motor Integration Workshop, ESF workshop,
Evolvability Symposium, 2 Italian Alife Workshop, Italian Psych. Meeting)
-
Turner’s
PhD transfer Report + PhD thesis
-
Marocco’s
PhD Thesis
|
