Author : Mehul Jain 1
Date of Publication :22nd August 2017
Abstract: Current voice-based digital assistants despite their claims of being intelligent, lack abilities that a true personal assistant must possess like extendable skill set, dynamic adaptation and high context awareness. In this paper, we highlight some design and implementation requirements that must be met in order for the development of next generation digital personal assistants and propose a general architectural backbone that can be used to make headway for such personalized speechoperated assistive technology. In particular, we confer about issues of extensibility of the skill set used by the digital assistant, hypothesis generation and evaluation, extensive user adaptation, and redundant representations handling in the design. Further, we briefly discuss the research and development directions that are undertaken to tackle challenges put by such a system. We then consider a scenario and illustrate the data flow in our architecture.
Reference :