READING REMEDIATION USING COMPUTER SPEECH RECOGNITION

Reading failure occurs in our schools at unacceptably high levels, causing serious and widespread social, political, and economic problems. This occurs despite the availability of effective instructional programs. Unfortunately, teachers often do not have the necessary training or time to implement these programs. In addition, it is sometimes not cost effective to use these programs with small or scattered student programs. This project will incorporate advanced speech recognition technology with the computer delivery of an effective, research-proven reading programs to effectively and economically teach basic reading skills to elementary redial and at-risk students. In Phase I, the first ten lessons were produced and evaluated with low- performing students using a speech recognizer trained by our staff for local dialect and similar students receiving "standard care." In a randomized trial, the treatment subjects improved significantly more than controls on phonemic segmentation. In Phase II, we will develop the complete 100 speech recognition remedial reading program for elementary students. The program will include a speech recognition engine that is sensitive to the dialect regions of the United States, complete multimedia presentation of instructional content, printed materials and all necessary hardware. The effectiveness of the program will be tested in a randomized control trial.