Quantifying Patterns of Sound Units in Languages Using Maximum Entropy

The frequencies of patterns of letters in words can be used to define pairwise interaction energies between letters using Jaynes’ principle of maximum entropy. By quantifying these interactions, overarching patterns in the English language can be studied [1,2]. We extend this framework to study patterns among combinations of sounds in words by investigating a phonological parametrization of texts in the English language, in addition to studying combinations of letters in words with the standard orthographic representation of these texts [1,2]. The benefit of this approach is that languages with different alphabets can be compared directly.

[1] G. J. Stephens and W. Bialek, Physical Review E 81, 066119 (2010) 

[2] A. Corral and M. G. Muro, Entropy 22, 179 (2020)