On Functional Load in Phonemics
Jørgen Rischel · 2008
Abstract If a linguistic text is looked upon as a unilinear sequence of phonemes, it is a rather simple procedure to measure the relative frequency of each phoneme. One sets up an overall inventory of phonemes and postulates that it is meaningful to characterize the use of the different phonemes as ‘drawn from a hat’. The overall frequencies are computed, and the probability of occurrence of each phoneme is expressed by its relative overall frequency. One can also focus on ‘items in context’ and compute the probabilities of the different phonemes occuring in a specific context or in a specific class of contexts. And finally, one can look upon the text as a ‘linear Markoff-process’ and interpret the probability of a specific sequence as the product of the subsequent transition probabilities. This in itself is a quite trivial procedure.