paper available

[Mary Hare]

The following paper is available by anonymous ftp from crl.ucsd.edu.


LEARNING AND MORPHOLOGICAL CHANGE

Mary Hare
Dept. of Psychology
Birkbeck College, University of London
hare at crl.ucsd.edu

Jeffrey Elman
Dept. of Cognitive Science
U. of California, San Diego
elman at crl.ucsd.edu

ABSTRACT:
This paper offers an account of change over time in English verb morphology,
based on a connectionist approach to how morphological knowledge is acquired
and used (Rumelhart and McClelland 1986, Plunkett and Marchman 1991, 1993).
A technique is first described that was developed for modeling historical
change in connectionist networks, then that technique is applied to model
English verb inflection as it developed from the highly complex past tense
system of Old English towards that of the modern language, with one predominant
regular pattern and a limited number of irregular forms.

The model relies on the fact that certain input-output mappings are easier than
others to learn in a connectionist network.  Highly frequent patterns, or those
that share phonological regularities with a number of others, are learned more
quickly and with lower error than low-frequency, highly irregular patterns
(Seidenberg and McClelland 1989).  A network is taught a data set representative
of the verb classes of Old English, but learning is stopped before errors
have been eliminated, and the output of this network is used as the teacher
for a new network.  As a result, the errors in the first network are passed
on to become part of the data set of the second.  As this sequence is repeated,
those patterns that are hardest to learn lead to the most errors, and over time
are 'regularized' to fit a more dominant pattern.

The results of the network simulations are highly consistent with the major
historical developments.  These results are predicted from well-understood
aspects of network dynamics, which therefore provide a rationale for the
shape of the attested changes.


*************************** To obtain a copy ************************


unix> ftp crl.ucsd.edu
Name: anonymous
Password: (type your email address)
ftp> cd pub/neuralnets
ftp> binary
ftp> get history.ps.Z
ftp> quit
unix> uncompress history.ps.Z
unix> lpr  history.ps (or what you normally do to print PostScript)