Challenging the Perceptual Relevance of Prosodic Breaks in Multilingual Spontaneous Speech. Corpora: C-ORAL-BRASIL / C-ORAL-ROM


Massimo Moneglia, University of Florence
Tommaso Raso, 2UFMG
Maryualê Malvessi-Mittmann, 2UFMG
Heliana Mello, 2UFMG

A Corpus of Brazilian Portuguese (BP) will join C-ORAL-ROM [1] adopting the same corpus design and prosodic annotation schema. The inter-rater agreement concerning the annotation of terminal and non terminal breaks by both experts and non experts is studied and compared with the early C-ORAL-ROM results [2]. Although the overall prominence of prosodic breaks is confirmed (K $>$ 0.80) the inter-rater agreement for terminals turns out satisfactory only for the experts (0.76). Moreover the annotation of non terminal breaks shows low reliability and suffers of language specific factors connected to the rhythmic structure of BP [3]. The paper focuses on the qualitative analysis of the language contexts types determining the low inter-rater agreement and highlights both language specific and general factors which interact with perceptual prominence of prosodic breaks in BP.