End-to-End Arabic Speech Recognition: A Review

Abdelhamid, Abdelaziz, Hamzah A Alsayadi, Islam Hegazy, Zaki Taha Fayed,


Automatic speech recognition (ASR) is a crucial field of science due to its massive applications that can be developed to help humans to improve their daily life tasks. Despite its long history, ASR remains an active and interesting research field in general and on Arabic language in particular. Arabic is one of the most widely spoken languages. However, current research is still limited on it due to its high variations and complex morphology. Therefore, this paper highlights the most recent techniques and key milestones of Arabic speech recognition to guide researchers who are interested in working on the Arabic language. There are many machine learning techniques applied in building ASR systems. For long time, hidden Markov models (HMMs)-Gaussian mixed models (GMMs) were standing as the best frameworks for ASR. However, in last decade, hybrid HMM-deep neural network (DNN) models and end-to-end deep learning models have been emerged as a breakthrough in improving the performance of ASR. End-to-end deep learning is distinguished as the most recent methodology in the field and represents the main focus of this review. Therefore, the proposed review discusses the most recent achievements of research on Arabic speech from the end-to-end methodology perspective. In addition, the currently available services and toolkits necessary for building end-to-end models are explained.

Other data

Title End-to-End Arabic Speech Recognition: A Review
Authors Abdelhamid, Abdelaziz ; Hamzah A Alsayadi ; Islam Hegazy ; Zaki Taha Fayed 
Affiliations Faculty of Computer and Information Sciences ; Faculty of Computer and Information Sciences ; Faculty of Computer and Information Sciences ; Faculty of Computer and Information Sciences 
Keywords LSTM;Automatic Speech Recognition (ASR), End-to-End;Deep learning;CTC;RNN;Attention-Based;HMM
Issue Date Sep-2020
Publisher Bibliotheca Alexandrina
Conference The 19th Conference on Language Engineering (ESOLEC’19)

Attached Files

File Description SizeFormat Existing users please Login
ArabicASRReview.pdf697.07 kBAdobe PDF    Request a copy
Recommend this item

Similar Items from Core Recommender Database

Google ScholarTM


Items in Ain Shams Scholar are protected by copyright, with all rights reserved, unless otherwise indicated.