Predicting Kurdish EFL university learners' oral reading fluency using support vector machine

Abstract EN

Investigating learners' English Oral Reading Fluency (ORF) in the contexts where English is used as a foreign language (EFL) or second language (ESL) has recently become a trending subject.

This study was carried out to predict the ORF of 100 Kurdish EFL university students of the English language department, college of Basic Education, university of Duhok, Iraq in 2020 during covid-19 using the support vector machine (SVM) technique.

This technique is one of the supervised machine learning techniques, and it is considered the most powerful algorithm in machine learning in terms of high accuracy; therefore, it was employed in this study.

Participants' ORF was measured by two experienced human raters using the four dimensions of the Multidimensional Fluency Scale (MDFS) including expression & volume, phrasing, smoothness, and pacing, which were used as the input variables to predict the ORF as an output.

Six kernels of the SVM were used in the prediction process.

The results indicated that the highest accuracy of testing result was obtained on the use of SVM Linear kernel with a value of 96.2%.

Confusion matrix was utilized to assess the outcomes of data classification.

The results of precision, recall, and FI-score were the highest for the SVM linear kernel and their values were the same for all performance metrics with a value of 96.1%.

Accordingly, it can be concluded that the performance of the SVM is considerably accurate in predicting the oral reading fluency.