Vowel systems demonstration links
This page contains links to the executable and its source code that
were used at the Evolang 2004 tutorial on computer modelling of
language. The code is far from flawless, but does allow for some
exploratory experiments. I will try to fix some of the more obvious
bugs if time permits.
The exexutable and the source can be downloaded for free, under the
condition that if any use of this material is made in a scientific
publication, a reference to one of my articles on the topic is included.
For general information:
de Boer, B. (1997)
Generating vowels in a population of agents. In P.
Husbands & I. Harvey (eds.) Proceedings
of the Fourth European
Conference on Artificial Life, Cambridge (MA) MIT Press, pp.
503–510
Bart de Boer (2000) Self organization in vowel systems, Journal of Phonetics 28 (4), 441–465
Bart de Boer (2001) The origins
of
vowel systems, Oxford: Oxford University Press
For information on changing populations:
Bart de Boer &
Paul Vogt (1999) Emergence of Speech Sounds in
Changing Populations In: Dario Floreano, Jean-Daniel Nicoud &
Francesco Mondada (eds.) Advances in
Artifical Life, Lecture Notes in
Artificial Intelligence 1674,
Berlin Springer Verlag, pp. 664–673
Some information on the simulation
This simulation was written as a demo version of the vowel systems
simulations that were used in the papers referred to above. It can be
used to execute the imitation game in its simplest form, but also to
execute the imitation game with changing populations or with two
spatially separated populations. The behavior of the experiments can be
changed by changing parameters (under the menu Parameters).
The synthesis parameters determine articulator- and acoustic noise.
Changing the value of these different kinds of noise. will change the
number of vowels that emerges.
The agent parameters determine two things. Most parameters determine
the ageing and learning behavior of the agents. The stepsizes determine
how quickly an agents approaches a freshly learned vowel target. The
other variable in the menu, lambda, determines the ratio between
possible horizontal distinctions and possible vertical distinctions.
The simulation parameters are some of the internal parameters (without
much cognitive significance), described in the papers. The "number of
cycles" is the number of games that is played when one chooses the menu
item Games/Multiple.
Under the item statistics, some measures of the simulation's behavior
can be found.
The distance parameters determine how the probabilities to interact
change with distance. For each agent-agent pair a weight is defined as
follows: wij = 1/(dist(i,j)+threshold)importance. The probability of an
agent i interacting with one
of the other agents j in the
population is then the weight wij divided by the sum over all weights wik,
where k<>i
The simulation presents four windows. The upper left window shows the
spatial distribution of the agents. The Right two windows show the
vowel systems of the agents from the two different population centres.
The lower left window shows textual information on the imitation games.
All agents have a letter associated with them, and these letters are
also used when plotting the agents vowels. One can get more information
about different vowels by clicking on the vowel system windows, but
this is not guaranteed to work, unfortunately.
Downloads
Windows executable (SpaceAge2.exe, 2.25 MB)
includes debug information.
Source code of microsoft visual C++ project (Zipped,
97K)