Reading Fluency Evaluation for Malaysian Primary School Children Using Feature Extraction Techniques in Speech Recognition

Abstract

It is no longer compulsory for lower primary school children to sit for examinations, they will now be evaluated through a more objective assessment, known as PBD (classroom-based assessment). Though it may seem to benefit both teachers and students in reducing stresses and anxiety over sitting for exams, but it would cause teachers to be burdened by the extremely high work-loads due to the frequent assessment. This could result in extreme pressure for the teachers which would eventually reach to the point of non-objectivity as well as inconsistency in students’ evaluation. Hence, this paper intends to investigate on the ways of easing the teachers by exploring the subject of speech recognition technique with the ability to evaluate students’ fluency based on a standardized set of criteria in reading assessment. Initial study conducted shows that the technique that is frequently used in speech recognition is Mel-Frequency Cestrum Coefficient (MFCC) method which was found to be able to extract the features needed in evaluating the fluency of a speech. Moreover, these extracted features are able to match with the criteria of a fluent reading which are pausing, phrasing, intonations, accuracy as well as pace of reading.