Files in this item



application/pdfFA20-ECE499-Thesis-Wang, Yuechen.pdf (301kB)Restricted to U of Illinois
(no description provided)PDF


Title:Analysis of phone-errors in reading disabled children
Author(s):Wang, Yuechen
Contributor(s):Allen, Jont
Subject(s):Reading Disability
Speech Perception
Abstract:Reading disability (RD) is common among children. Negative effects typically persist into the adulthood. We show that RD children's errors are highly idiosyncratic. Thus, specific treatments must be made based on their specific errors. In this study we looked deeper than previous experiments, providing a new approach in analyzing the data. The experiments involved eleven RD subjects and six reading control (RC) subjects. Two tasks were performed: the Syllable-Confusion Oddball (SCO) task which can determine which phones cause errors, and the Nonsense Syllable Confusion Matrix (NSCM) task which provides an error map identifying the phone perception confusions for each subject. A seventh subject Average Normal (AN) is created in the control group who has the average performance of the six RC subjects. AN serves as the control subject inside the RC group, allowing us to identify the general confusion patterns among the normal children, and to compare the RD and RC groups. We generated a confusion matrix for each subject based on the NSCM task and calculated the angles between the column vectors for the same sound. The smaller the angle, the less difference in phone perception confusions between two subjects for one sound. To find out the confusion patterns of both RC and RD subjects, we clustered the off-diagonal errors to form confusion groups. We found out that RC subjects and AN subject had a significant overlap in confusion groups, which strongly suggested that six RC subjects are sufficient to draw the pattern (that of the AN subject) of confusion groups of RC subjects. However, RD subjects had much higher error rates and nearly random confusion groups compared with the RC subjects. Each RD subject has unique confusion pairs, and has either much higher error rates, or unique confusions. The results support the hypothesis that RD subjects are idiosyncratic. Thus, a larger number of subjects will not reveal distinct patterns, since every subject has unique errors (there are no patterns). We believe that the confusion groups we found should be helpful in the subsequent treatment of RD children.
Issue Date:2020-12
Date Available in IDEALS:2021-01-04

This item appears in the following Collection(s)

Item Statistics