Improving the Memory of Intelligent Personal Assistants

Document Type

Student Presentation

Presentation Date


Faculty Sponsor

Casey Kennington


Intelligent Personal Assistants (IPAs) of today (e.g., Google’s Assistant, Apple’s Siri, and Microsoft’s Cortana) are rapidly transforming how humans interact with computers. However, these systems commonly behave like amnesiacs; they often do not remember what the user told them seconds ago, which inhibits their ability to learn new concepts over an extended period of time. This research project proposes a model that will enhance the memory capabilities of IPAs by describing an incrementally flexible dialogue system architecture. The model would enable an IPA to form long-term concepts and relationships from natural language input without the use of a predefined schema. Not only would these concepts improve an IPA’s response to certain types of queries, user interaction would feel more personal and natural.

This document is currently not available here.