Axe 1 - Workshop "Speech variation: Impact on perception and comprehension" - 17/18 janvier 2020

Quand ? Du 17-01-2020 à 09:00
au 18-01-2020 à 12:30
Où ? Salle plate 031 - RdC MSHS
S'adresser à
Participants Katie Drager - Department of Linguistics, University of Hawai, Mānoa;
Sophie Dufour- CNRS, LPP, Aix en Provence.
Bronwen Evans- Speech, Hearing & Phonetic Sciences, UCL, London, UK.
Michel Hoen, Oticon Medical, Vallauris.
Sven Mathys, Department of Psychology, Université de York, UK.
James McQueen – Max Planck IInstitute, Nijmegen, NL
Julien Meyer - CNRS, Gipsa Lab, Grenoble.
Arthur Samuel, Cognitive Science, Psychology, Stony Brook University, USA.
Ajouter un événement au calendrier vCal
iCal

In the real world, we hear speech that is often deviating from the standard pronunciation. The first difficulty with speech is that each production is unique. Dictionaries give the canonical forms of words, yet in connected speech the pronunciation of these words varies a lot. Pronunciations with fewer segments or syllables, for instance, are frequent in many languages. In French, for example, many words have a schwa /médecin/, and this schwa is often not produced (e.g., /méd’cin/). The pronunciation of many words also often depends on contextual constraints. For instance, in French, as in many languages, word-final sounds can be produced differently depending on the phonological properties of the following word; For example sac /sak/ “bag” is being realized as [sag] in sac de graines “bag of seeds” because of voice assimilation. Moreover variability in speech productions also includes factors that concern the talker such as talker’s dialect, gender, native language, age, physical stature but also speaking rate, speaking style and emotional state. Language acquisition and social interactions, to be successful, require the ability to compensate for this substantial variability. It is particularly impressive that indeed most of the time the chain of processes involved in speech processing reach their goal: the message is understood. This robustness of speech perception and language processing in the face of the variability of acoustic signals is a central puzzles in the study of speech perception and modelisation, however very few has been done to evaluate how it impacts comprehension.

How such variability of speech production modulates perception and even more crucially comprehension? The workshop will propose to tackle this issue from a cognitive science perspective.

During the last century, the study of language has been conducted under the influence of very strong theoretical propositions such as the structuralism (as carried by Jakobson and Saussure for example) or the modularity of mind (as carried by Fodor). These frameworks encouraged the study of language processing in isolation from other cognitive domains or sensory capacities. Recently however, a more integrated approach for language has emerged. As a highly integrated cognitive function, speech comprehension indeed relies on a chain of interconnected processing stages that starts with auditory encoding, but also needs the recollection of lexical and semantic information stored in memory and culminates with the understanding of the message. Ultimately, the success of the complete procedure depends on the success of each step, with the potential help of compensation mechanisms between the different stages.

During this workshop we will bring together researchers that tackle this issue using different distortion of speech and different perspectives. Our aim is to promote interactions and future collaborations on this very fruitful topic of cognitive science and a real challenge for modeling and engineering.