Dialect classification using acoustic and linguistic features in Arabic speech

<span lang="EN-US">Speech dialects refer to linguistic and pronunciation variations in the speech of same language. Automatic dialect classification requires considerable acoustic differences between different categories speech. This paper proposes a model composed combination classifiers for Arabic by utilizing both features spontaneous The comprises an ensemble focusing on frequency ranges within short-term spectral features, as well classifier ‘i-vector’, whilst use extracted transformer models pre-trained large text datasets. It has been shown that proposed fusion multiple achieves accuracy 82.44% identification task five dialects. represents highest reported dataset, despite relative simplicity model, its applicability relevance tasks. </span>