Sixth European Conference on Speech Communication and Technology
(EUROSPEECH'99)

Budapest, Hungary
September 5-9, 1999

Prosody as a Distinctive Feature for the Discrimination of Arabic Dialects

Melissa Barkat (1), John Ohala (2), François Pellegrino (1)

(1) Laboratoire DDL-ISH, Lyon, France (2) Phonology Laboratory, University of Berkeley, CA, USA

The aim of the work to be reported here is to explore the utility of prosodic information in language identification and discrimination tasks. The purpose of this study is to see whether prosodic patterns can be considered as reliable acoustic cues for the discrimination of Arabic dialects by investigating, via a perceptual experiment, if listeners are successful in identifying the Arabic dialect used by a speaker when they only have access to fundamental frequency, amplitude and some rhythmic characteristics of the original voice signal. Results show that prosodic cues alone can distinguish between dialect pairs, since native Arabic listeners are significantly more successful in identifying the Arabic dialectal varieties both in their natural and synthesized forms and that listeners’ identification rate are higher for the discrimination of their own dialectal variety when presented under its processed form. This perceptual study must be regarded as a first step towards the determination of a set of reliable cues for the Automatic Identification of Arabic Dialects.


Full Paper (PDF)

Bibliographic reference.  Barkat, Melissa / Ohala, John / Pellegrino, François (1999): "Prosody as a distinctive feature for the discrimination of arabic dialects", In EUROSPEECH'99, 395-398.