Goal-Oriented Modelling for Virtual Assistants

Jonathan Leung, Zhiqi Shen, Chunyan Miao

Abstract: Virtual assistants are used in a wide variety of environments by different types of users. Giving users the ability to build and customize virtual assistants' skills and capabilities would enable them to create virtual assistants that can fit the needs of different scenarios. We propose a model for virtual assistants, based on Goal Net, with the aim of empowering users without programming experience to personalize and customize their virtual assistants. Goal Net separates the design of agent mental models from their low-level implementation. Developers contribute to a library of functions which can be used designers to develop functionality for their virtual assistants. The Multi- Agent Development Environment (MADE) is a graphical tool for creating Goal Net agents and allows users to easily deploy their agents for usage without the need to compile code. A case study is performed to show how Goal Net can be used to develop virtual assistant skills. The proposed model provides a foundation for future work, which would involve human computer interaction and natural language processing.