Automatic pitch accent annotation

Grazyna Demenko, Magdalena Oles?kowicz-Popiel

Based on a non-expressive speech data corpus of a few hundreds of utterances from 80 speakers a description and annotation of pitch accent shape have been proposed. Phrases were transcribed automatically on segmental level and lexical accents were annotated in accordance with expert rules using self-developed software. Pitch shape annotation was based on F0 changes on accented and postaccented syllables and parameters related to the register defined on a frequency physical scale and F0 change range. The indicated accent structures were verified manually and then evaluated statistically. Statistics have shown that accent in Polish is most frequently realized through F0 level changes between accented and postaccented syllable.

DOI: 10.21437/SpeechProsody.2016-16

Cite as

Demenko, G., Oles?kowicz-Popiel, M. (2016) Automatic pitch accent annotation. Proc. Speech Prosody 2016, 74-78.

author={Grazyna Demenko and Magdalena Oles?kowicz-Popiel},
title={Automatic pitch accent annotation},
booktitle={Speech Prosody 2016},