Title: Sound model for dialogue profiling
Authors: D. López De Luise; R. Azor
Addresses: CI2S Lab, Pringles 10, 2nd FL. Bs As, C1183ADB, Argentina ' Universidad de Mendoza-FI, Pasaje Descotte 750, Mendoza, 5500 Argentina
Abstract: Zipfian behaviour has been defined as a way to predict word distributions in natural language for written texts. It keeps valid for any Language. This type of simple models opens a door for practical applications for understanding how the brain works and produce sentences and dialogues in a very efficient and robust way. Currently, many applications make profit of advances in understanding linguistics and psycholinguistics. But those advances have not a similar counterpart in speech and utterances modelling fields. This paper presents a fractal model for oral productions in Spanish. The results of the analysis show that samples from a sound library with Spanish voices can be assimilated to a specific fractal distribution. Disregarding which is the speaker, the trend keeps valid for every sample. This finding shows that also spoken language performs.
Keywords: natural language processing; dialogue processing; fractals; Zipfian behaviour; language metrics; dragon fractals; linguistic reasoning; reasoning modelling; language production; speech processing.
International Journal of Advanced Intelligence Paradigms, 2017 Vol.9 No.5/6, pp.623 - 640
Received: 27 May 2015
Accepted: 21 Jul 2015
Published online: 27 Nov 2017 *