24.4 MOVING SPEECH TECHNOLOGY METHODS OUT OF THE LABORATORY: PRACTICAL CHALLENGES AND CLINICAL TRANSLATION OPPORTUNITIES FOR PSYCHIATRY

摘要

Psychiatric patients, such as those suffering from depression or schizophrenia, often need to be monitored with frequent clinical interviews by trained professionals so as to avoid costly emergency care and unfortunate events (e.g., suicide attempts). Technological advances in the form of smart devices offer a mobile platform through which to provide the effective and affordable monitoring of clinical events. Novel speech technologies offer promise of increased efficiency, sensitivity and objectivity by the implementation of automatic speech recognition and natural language processing methods to facilitate in the tracking of the clinical state of psychiatric outpatients longitudinally and, when appropriate, alerting clinical staff to contact that patient. However, thus far most research that has leveraged such technology has been conducted within controlled laboratory or clinical environments, and as such it remains unknown how robust such methods would be when the data collection is in uncontrolled settings and controlled by the participants themselves. Yet if these methods are to truly have clinical translation value then they must be demonstrated to be user-engineered to nurture participation and to be tolerated by participants despite frequent use, and that the resulting behavioral responses - notably voice - that are collected in uncontrolled settings remain interpretable by speech recognition and natural language processing methods.