Automated Assessment of Aphasic Speech Using Discrete Speech Recognition Systems

摘要

The need for computer-based aphasia treatment programs (CBATP) is increasing as viable options are sought for optimizing the quality, quantity and accessibility of therapy while also reducing its cost. A challenge in the development of CBATPs has been automated assessment of spoken responses. This feature would allow for provision of feedback to the patient and/or performance data to the speech-language pathologist. In this poster, we present data comparing the automated judgments of response accuracy of several systems with that of experienced clinicians. Patient data for this project was downloaded from the AphasiaBank database (McWhinney, Fromm, Forbes, & Holland, 2011).