Linguistic Applications

[Teresa Satterfield]

Andrew:
I've used Swarm to model linguistic phenomena (acquisition of syntactic
features in creole grammars,to be precise) and to simulate syntactic
patterns which emerge from language contact situations.  I am currently
working (with a more accomplished Swarm programmer) to refine the basic
model which was presented at the LSA (Linguistic Society) Annual Meeting.
Abstract below:

Universal similarities found across diverse creole grammars are often
attributed, most notably in terms of syntax, to either i) social
conditions arising from language contact or ii) innate/genetic language
mechanisms.  Creole grammars are typically defined as those in which
simplified languages are invented as a means of communication in certain
contact scenarios by groups of adult speakers who have differing native
languages.  These adults subsequently pass on the conventionalized "new"
language to their children, for whom it becomes a mother tongue. In this
project, I offer a plausible account for the emergence of syntactic
similarities in principle as based on a computational language learning
model which couples general aspects of language contact dynamics with the
notion of linguistic innateness; the latter as largely outlined in the
Language Bioprogram Hypothesis (Bickerton 1984) and Chomsky (1981, 1986).

The approach taken will be to first create a population of agents,' based
on the historical components of the Latin American slave plantation.  Each
agent as slave or planter, is endowed with invariable linguistic
principles along with certain options/parameters which vary according to
the language present.  Each agent also possesses a set of rules which
stipulate the environmental conditions required for the acquisition of the
relevant options.  The agents are subsequently placed together in an
artificially constructed society consisting of a finite environment space
and finite natural resources.  The dynamics of the population are driven
by the agents' varying needs for multiple resources and their migration
from place to place within the artificial world in order to obtain
resources.  Native language attributes such as vocabulary and syntactic
structures are pre-assigned to the starting population of agents, but
subsequent generations of agents come with the innate capacity for
linguistic knowledge, however they acquire their observeable language
characteristics solely via social aspects and social contact situations
(i.e., through wealth, social status, the dynamics of trade and migration)
with other agents.  The model will be created using SWARM, a computerized
modeling system optimized for the simulation of complex dynamical systems
and will be refined based on the updated SWARM software and computer
processors.

>From a broader theoretical perspective, the proposed model offers insights
into the nature of linguistic input data and markedness theory (the notion
that some linguistic forms are more universal, more "natural" than other
forms).  Experimental models in language genesis can provide critical
insights into the particular computations that underlie the language
acquisition capacity in general.  More specifically, the current
investigation seeks to address the nature of the creolization process.
There are several hypotheses for how properties of languages and/or
social contact each come to bear on the resulting creole language, however
until recently there has been no reliable method for testing these ideas.
Computer implementations such as the proposed model of possible strategies
can illustrate the overall behavior of the system.  The simulation of
machine-based language "learners" allows a rigorous testing of data and
re-creation of historical occurrences in order to indicate whether either
or both types of hypotheses really may be effective in solving an
important problem in the acquisition of languages.  

In sum, by building computational language systems we often can uncover
difficult aspects of a problem and possible solutions which were not
realized upon first consideration.   As I claim, the creolization model
provides just this type of viable response in terms of advancing a new
"sociogenetic" solution for the emergence of  historical creole languages. 
 

Teresa Satterfield
University of Michigan
Program in Linguistics
address@hidden
www.umich.edu/~tsatter




                  ==================================
   Swarm-Modelling is for discussion of Simulation and Modelling techniques
   esp. using Swarm.  For list administration needs (esp. [un]subscribing),
   please send a message to <address@hidden> with "help" in the
   body of the message.
                  ==================================