Training a confidence measure for a reading tutor that listens

One issue in a Reading Tutor that listens is to determine which words the student read correctly. We describe a confidence mea- sure that uses a variety of features to estimate the probability that a word was read correctly. We trained two decision tree classifiers. The first classifier tries to fix insertion and substitu- tion errors made by the speech decoder, while the second clas- sifier tries to fix deletion errors. By applying the two classifiers together, we achieved a relative reduction in false alarm rate by 25.89% while holding the miscue detection rate constant.