Kalender Details

Vortragende Person/Vortragende Personen:
Antje Schweitzer (Institut für Maschinelle Sprachverarbeitung, Universität Stuttgart)

According to exemplar theory, phonological categories correspond to accumulations of similar perceptual tokens. Consequently it should be possible to acquire these categories based on their parameter distributions in a perceptual space. In this talk, I will investigate parameter distributions of adult intonation categories in a phonetic space. I will argue that the perceptual space can be approximated by finding appropriate weights for the phonetic dimensions. Using these weights, I will explore clustering as a means to detect accumulations of similar tokens, and evaluate how well they correspond to actual intonation categories. It will be shown that the weights indeed improve the detection of categories via clustering, but that the detection of intonation categories is more difficult than the detection of vowel categories for instance.