Comparative Analysis of Prosodic Characteristics of Spontaneous and Synthesized Speech (Based on Kazakh and English Ted Talks Video Materials)

Abstract

The study aims to conduct an instrumental-comparative analysis of the prosodic characteristics of spontaneous (based on TED Talks materials) and synthesized speech in Kazakh and English. The paper examines existing prosody research approaches and an acoustic analysis of key prosodic parameters (pitch frequency, intensity, and tempo) for spontaneous and synthesized speech types. For the comparative analysis, a corpus was developed, containing 10 speech excerpts drawn from TED Talks each in Kazakh and English, which were then transcribed and converted into audio files using modern speech synthesis systems. The acoustic analysis was conducted using PRAAT software and own proprietary software, ProAG-2025 (protected document No. 58731, dated May 27, 2025). This article formulates a hypothesis that spontaneous speech is characterized by greater variability in prosodic features, while synthesized speech differs from natural speech in acoustic and prosodic features. The instrumental analysis results confirm that synthesized speech, despite its structural conformity, retains a set of parameters that allow it to be reliably differentiated from spontaneous speech in increased amplitude uniformity and frequency contours, the absence of stochastic variations, and a simplified rhythmic-pause pattern. The obtained data are of practical significance for the further improvement of speech synthesis algorithms, increasing the degree of naturalness, and optimizing the communicative effectiveness of media applications.

Keywords: spontaneous speech, synthesized speech, prosody, acoustic parameters, tonality, pitch frequency.