9th Annual Conference of the International Speech Communication Association

Brisbane, Australia
September 22-26, 2008

Testing a Large Corpus of Natural Standard Arabic for Rhythm Class

Liz Dockendorf, Dalal Almubayei, Matthew Benton

University of Texas at Arlington, USA

Previous studies using acoustic correlates to measure speech rhythm have used small samples of audio and a limited number of speakers. Few have included standard Arabic in the analysis. This study uses Arabic news broadcast along with data output from an automatic speech recognizer time-aligned transcript to test over 50 minutes of speech by 46 speakers. The results show that Arabic, like English and Chinese news broadcasts, exhibits a range of individual speaker rhythm, but its average clearly distinguishes it from the other two on the rhythm continuum. A large number of speakers also makes the average more reliable.

Full Paper

Bibliographic reference.  Dockendorf, Liz / Almubayei, Dalal / Benton, Matthew (2008): "Testing a large corpus of natural standard Arabic for rhythm class", In INTERSPEECH-2008, 771.