Mohd Yusof, Shahrul Azmi and Atanda, Abdulwahab F. and Hariharan, M. (2013) A review of Yorùbá Automatic Speech Recognition. In: 3rd International Conference on System Engineering and Technology, 19 - 20 Aug. 2013, Shah Alam, Malaysia.
PDF
Restricted to Repository staff only Download (792kB) |
Abstract
Automatic Speech Recognition (ASR) has recorded appreciable progress both in technology and application.Despite this progress, there still exist wide performance gap between human speech recognition (HSR) and ASR which has inhibited its full adoption in real life situation.A brief review of research progress on Yorùbá Automatic Speech Recognition (ASR) is presented in this paper focusing of variability as factor contributing to performance gap between HSR and ASR with a view of x-raying the advances recorded, major obstacles, and chart a way forward for development of ASR for Yorùbá that is comparable to those of other tone languages and of developed nations.This is done through extensive surveys of literatures on ASR with focus on Yorùbá.Though appreciable progress has been recorded in advancement of ASR in the developed world, reverse is the case for most of the developing nations especially those of Africa.Yorùbá like most of languages in Africa lacks both human and materials resources needed for the development of functional ASR system much less taking advantage of its potentials benefits. Results reveal that attaining an ultimate goal of ASR performance comparable to human level requires deep understanding of variability factors.
Item Type: | Conference or Workshop Item (Paper) |
---|---|
Uncontrolled Keywords: | Automatic Speech Recognition, Robust ASR, Variability in ASR, Yorùbá speech processing |
Subjects: | Q Science > QA Mathematics > QA76 Computer software |
Divisions: | College of Arts and Sciences |
Depositing User: | Dr. Shahrul Azmi Mohd. Yusof |
Date Deposited: | 21 Apr 2014 23:58 |
Last Modified: | 02 Nov 2020 02:22 |
URI: | https://repo.uum.edu.my/id/eprint/10737 |
Actions (login required)
View Item |