Task Characteristics of a Computerized TTS Dictation Test and Diagnostic Interpretability of Qualitative Error Analysis

Abstract

The present study attempted to explore the overall efficacy of a computerized text-to-speech dictation test in terms of the task characteristics and the diagnostic interpretability of the test results through automatized error analysis. The computerized dictation test was designed to be conducted in the following four stages: first, subjects listened to the entire passage; second, they typed what they heard; third, they revised their answers (in two different modes: with and without feedback on the previous input); fourth, the computer displayed their scores. Based on the data analysis conducted by the automated scoring program, the research findings revealed the following: 1) the test task characteristics (the absence and presence of immediate feedback on students’ answers and the test item length) of computerized dictation test exerted a significant impact on test taking strategies and subsequent performance; 2) the function word errors were revised less frequently than the content word errors, revealing a problem with students’ comprehension of unstressed phonological sandhi phenomena; 3) the qualitative error analysis implied the diagnostic potential of the computerized dictation test; and 4) the computer-based dictation test was preferable to the conventional paper-based version.