A Neuron-Computational Model of Syllabification applied in Persian and Balochi languages
Dr. Omid Reza Taheri
Researcher in Philosophy of Mind & Cognitive Science
26 May 2013
This model is an insight into both philosophy of mind and language. The reason for designing such model is to analyze the unobservable process of a small part of human language, i.e. syllabification. The achievement behind this designed model is based on studies of several domains, such as philosophy of language and Linguistics, Neurophysiology and Genetics, Philosophy of mind and Symbolic Artificial Intelligence, and Connectionism and Dynamicism.
It is important to mention that the designed model inspired from the works on theories of language (especially phonology) proposed by Noam Chomsky, John Goldsmith and Bernard Laks. The model consists of 15 units (neurons) which connected to each other through the connection weights. Each assigned initial weights. The phonological system of both Persian and Balochi language are classified into different levels of sonority, starting from lowest consonants, ending to most sonorant vowels.
For each I have assigned certain values based on their sonority hierarchy. As it is mentioned earlier, the phonemes assigned certain values according to their sonority levels. The network of our syllable system provides and makes the proper harmonization among the units (neurons). Each unit inherently possesses a rule which is universal for all units. The interrelation between units occurs through a simple equation which provides the learning process. There are two important parameters in the program. The first is the sum of consonants values and the second is the value that analyzed for the threshold of each units.
Each unit consists of two rules: 1. Cognizer 2. Recognizer For each connection between units, I have assigned an initial value (weight or synapse). Finally, there is a value assigned as a learning rate. The summed value possesses the consonant segments. The cognizer consists of certain rule. When it receives or cognizes certain phonemes as an input, it matches it with the recognizer part of the program (that is another rule). If it finds that received phoneme matches with the existing one, then the program accepts it, and no need to be trained and learned by the program. But, if the inputs or received phonemes would not be recognized by recognizer’s rule, then it would be transferred to the learning part of the program where learning rate would increase or decrease the received value (sonority or energy) of the phoneme and then make the correct and proper connection weight between two related units or neurons.
Through such mechanism and harmonization the program syllabify the receiving strings of phonemes. The program receives all phonemes in parallel manner and gives its outputs in parallel. The program has been tested 3500 words for Persian and 2250 words for Balochi languages. The results were incredible for both languages. It has shown the perfect results of syllabification for both languages. I have concluded that there are two factors which underlie the syntactical structure of phonological system and the way they changed.
The first is related to the innate structure, much concerned with the deep predispositive structure of the syntactical structure of the certain phonology of language, and the second is related to the environmental issues and the way learning occurs within the certain socio-cultural systems.
I would like to invite the researchers who are involved in this area to have their ideas and comments.
© www.onlinethinker.org / All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means without referring to the source (TECODH). The views, conclusions, and suggestions expressed in this article/seminar are those of the author/presenter and do not necessarily reflect the views of TECODH.