Principles of Building a Prosodic Annotation System Based on Parametric Data

Principles of Building a Prosodic Annotation System Based on Parametric Data

Goncharova O.V. (ISP RAS, Moscow, Russia; RUDN, Moscow, Russia)

Abstract

The paper addresses the development of a reproducible prosodic annotation system that can capture both linguistically relevant categories and the gradient variability of the speech signal. A review of existing categorical and continuous approaches identifies recurring limitations, including annotator subjectivity, incomplete coverage of prosodic parameters, and limited comparability across speakers. To mitigate these issues, the study proposes a methodology grounded in comprehensive parametric modeling of speech. Its core component is a normalization procedure for the primary acoustic correlates of prosody-fundamental frequency (F0), intensity, and duration-mapped onto a unified speaker-relative scale (0–100) and subsequently discretized via thresholding into perceptually motivated levels. The proposed framework implements a two-layer representation: a lower layer comprising numerical measurements and derived dynamic metrics, and an upper layer represented by sequences of discrete level labels. The system is formalized and amenable to automation, enabling cross-speaker and cross-situational comparability while preserving the signal’s absolute characteristics. The paper details the methodological principles and implementation workflow, provides examples of output tables and graphical visualizations, and discusses the remaining limitations of the approach.

Keywords

prosodic transcription; Russian; intonation; normalization; fundamental frequency (F0); intensity; duration; syllable segmentation; threshold-based categorization; comparative analysis; speech signal.

Edition

Proceedings of the Institute for System Programming, vol. 38, issue 6, part 2, 2026, pp. 285-302

ISSN 2220-6426 (Online), ISSN 2079-8156 (Print).

DOI: 10.15514/ISPRAS-2026-38(6)-38

For citation

Goncharova O.V. Principles of Building a Prosodic Annotation System Based on Parametric Data. Proceedings of the Institute for System Programming, vol. 38, issue 6, part 2, 2026, pp. 285-302 DOI: 10.15514/ISPRAS-2026-38(6)-38.

Full text of the paper in pdf (in Russian) Back to the contents of the volume