摘要

This paper presents the Algerian Arabic Speech Database (ALGASD), a Modern Standard Arabic (MSA) speech corpus composed of utterances pronounced by 300 Algerian native speakers selected from eleven regions of Algeria. One of the objectives of this corpus design is to be representative of the regional accents of MSA spoken in Algeria. Useful information related to the speakers, such as gender, age, and education level, is provided. This paper also reports the results of the Automatic Speech Recognition (ASR) application of the corpus and outlines an original global monophone recognition model designed to handle linguistic variability. The global phone recognition rate for this ASR reference system is satisfactory and may constitute a useful baseline ASR system dedicated to MSA.

  • 出版日期2010-12