VIARS: An Intelligent Voice Agent to Prevent Inappropriate Content Display for Restricted Users

Abstract

The wide usage of voice intelligent assistants (VIAs), such as Siri, brings lots of convenience to our daily life. However, there are also kinds of issues related to VIAs displaying inappropriate contents to restricted users such as children. To address this problem, we developed a system that detects the inappropriate contents and rates them with defined ranks. In this manner, a threshold or suggestion could be given in order to avoid displaying inappropriate contents to restricted users. This evaluation system is called Voice Intelligent Aid Rating System (VIARS). With a developed VIARS prototype, we further tested its effectiveness with a designed experiment that consisted of several tasks. As a result, VIARS’s intelligent rating results were proven to be highly consistent with the experts’ manual rating results, indicating the reliability of the VIARS classification mechanism. It is also notable that when compared with experts’ manual rating, VIARS’s automatic rating mechanism showed significant advantages in response time.