CNRS

Rechercher




Accueil > Formation > SRPP > Programme 2017-2018

SRPP du Vendredi 11 Mai 2018

Yoshinori Sagisaka (Waseda University, Tokyo)

Titre : Computational description and control of sentiment information embedded in speech

Résumé :

In this talk, recent research activities are introduced in sentiment correlation analyses among speech, language and color studied at our research laboratory. Following the findings of fundamental correlations between communicative speech prosody and its impressions expressed by language, communicative F0 pattern is calculated. Using Japanese sentences consisting of adverbs showing magnitude, adjectives and final particles, the possibility of F0 control for communicative speech synthesis is shown. The other experimental trial is also introduced to show that this sentiment mapping between impression (language) and prosody (speech) can also be observed between speech and color. These two research studies indicate a new paradigm of cross-modal computational modeling of sentiment information processing between speech, language and image.