A Novel Skill Model Supporting Complex and Reliable Interactions between Users, VPA and External Services

Xizhe Zhang, Demin Yu, Min Liu, Zhongjie Wang

Abstract: With the development of AI and dialog technologies, Virtual Personal Assistants (VPAs) have been used on various types of mobile and edge devices for convenient man-machine interaction. VPA's capabilities are commonly called "skills". VPA can use skills to get users intent and provide corresponding services such as playing music. However, the skills supported by existing VPAs have many shortcomings. First of all, skills cannot support complex interactions between users and VPAs. The interaction process is often just a simple process. Secondly, many skills are unreliable because they frequently terminate execution when they receive unexpected input. Moreover, the services supported by current skills are very limited that cannot meet the diverse needs of users. In order to solve these problems, this paper proposes a novel skill model named CRSModel to support complex and reliable interactions between users, VPA and external services. In order to improve the complexity of interactions between users and VPA, CRSModel adds many configurations that are helpful to support complex interactions, and appends the dialogue flow processing mechanism to the execution process that can let users change the dialog flow. In terms of reliability, CRSModel's execution process can capture unexpected input and guide the user to input an expected value. Finally, in order to meet the diverse needs of users, CRSModel can support external services that satisfy the REST protocol, greatly increasing the variety of services.