LEC talk: Fermin Moscoso del Prado Martin

I introduce a new tool for the macroscopic description of human language. A language, as represented by a corpus of text, can be described by its symbolic periodogram. This is an objective tool, analogous to the spectrograms commonly used for speech analysis. I will show that, across many languages, periodograms reveal a strikingly universal ‘shape’. Overall, the pattern consists of a tendency to avoid repetition of structures in the very short time scales, a tendency for repetitions at the very long scales, and a neutral regime in between those two. I will discuss how such statistical universal relates to a large number of phenomena that have been described in psycholinguistic research. Despite the universality of the general pattern, subtle differences also reveal particularities of individual languages. These differences demonstrate a long-held –but unproven– hypothesis: As the syntactic structure in a language becomes less complex, its morphological structure becomes more complex, as would be predicted if total amount of linguistic structure remains fairly constant across languages.

References:

Moscoso del Prado, F. (in press) The universal `shape’ of human languages: spectral analysis beyond speech. PLoS One.
[preprint available from http://www.moscosodelprado.net/docs/fourier-plos.pdf]