Generating prosody by superposing multi-parametric overlapping contours

Holm, Bleicke; Bailly, Gérard

doi:10.21437/icslp.2000-513

Cited by 10 publications

(3 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Few attempts have been reported to learn automatically the actual shapes of embedded contours. Holm et al [15,16] propose an analysis-by-synthesis method for decomposing an f0curve under high-level constraints ( §3.2).…”

Section: Contour Shapesmentioning

confidence: 99%

Automatic generation of prosody: comparing two superpositional systems

Raidt,

Bailly,

Holm

et al. 2004

Speech Prosody 2004

View full text Add to dashboard Cite

We face many options when designing a system that automatically generates prosody from linguistic and paralinguistic information. The literature provides several candidate phonetic models, phonological models and mapping tools to actually implement the system. We detail here some dimensions along which these models have to be compared. We show also that systems employing quite similar phonetic models can still have radically different approaches. We present results of a first evaluation comparing two systems using a superpositional model of melody on a common multilingual prosodic database of spoken math formulae. We conclude that prosodic models and intonation theories could certainly benefit from well-defined tasks and fair benchmarks.

show abstract

Section: Contour Shapesmentioning

confidence: 99%

Automatic generation of prosody: comparing two superpositional systems

Raidt,

Bailly,

Holm

et al. 2004

Speech Prosody 2004

View full text Add to dashboard Cite

show abstract

“…General-purpose contour generators have been developed in order to be able to generate a coherent family of contours given only their scope. These contour generators are actually implemented as simple feedforward neural networks [13] receiving as input linear ramps giving the absolute and relative distance of the current syllable from the closest landmarks and delivering as output the prosodic characteristics for the current syllable (see Figure 1). Each network have very few parameters -typically 4 input, 15 hidden and 4 output units = 4*(15+1)+15*(4+1) = 139 parameters -to be compared to the thousands parameters necessary to learn a "blind" mapping between phonological inputs and prosodic parameters such as in [6,24].…”

Section: Contour Generatorsmentioning

confidence: 99%

Learning the hidden structure of intonation: implementing various functions of prosody

Holm,

Bailly

2002

Speech Prosody 2002

View full text Add to dashboard Cite

This paper introduces a new model-constrained, data-driven method to generate prosody from metalinguistic information. We refer here to the general ability of intonation to demarcate speech units and convey information about the propositional and interactional functions of these units within the discourse. Our strong hypothesis are that (1) these functions are directly implemented as prototypical prosodic contours that are coextensive to the unit(s) they apply to, (2) the prosody of the message is obtained by superposing and adding all the contributing contours [2]. We describe here an analysis-bysynthesis scheme that consists in both identifying these prototypical contours and separating out their contributions in the prosodic contours of the training data. The scheme is applied to databases designed to evidence various functions of intonation. Experimental results show that the model generates faithful prosodic contours with very few prototypical movements.

show abstract

“…Both models simulated efficiently the utterances of the original corpus by applying the first three principles. Recently, Holm and Bailly [15] have applied this to a corpus of read mathematical formulae, in utterances where the segmentation function (mathematical symbols) exactly coincides with the semantic content, and where this content can occur at whatever level. Their results confirm the independence hypothesis.…”

Section: 4 Principlementioning

confidence: 99%

A Gestalt morphology of prosody directed by functions: the example of a step by step model developed at ICP

Aubergé

2002

Speech Prosody 2002

View full text Add to dashboard Cite

This paper proposes a point of view consisting i n showing prosody as an emergent form perceived via Gestalt processing. Contours carry some function values at different levels through the superposition of independent contours. In this view, tonal phonology can be interpreted as a bottom-up sub-processing approach, and global form modelling as supported by a top-down approach. The model can integrate a large scale of functions from the linguistic domain to the expression of emotion. The founding principles of the model are explained and the most significant steps developed at ICP are recalled to illustrate this theoretical background.

show abstract

Generating prosody by superposing multi-parametric overlapping contours

Cited by 10 publications

References 8 publications

Automatic generation of prosody: comparing two superpositional systems

Automatic generation of prosody: comparing two superpositional systems

Learning the hidden structure of intonation: implementing various functions of prosody

A Gestalt morphology of prosody directed by functions: the example of a step by step model developed at ICP

Contact Info

Product

Resources

About