A procedure for automatic adaptation to the user in intelligent virtual agents

Paper i proceeding, 2019

This paper describes procedures for improving the functionality of task-oriented intelligent virtual agents (IVAs) and carrying out adaptation to the user in such agents. The functionality improvement is based on automatic extension of input patterns (rules), using a database of interchangeable phrases. This extension results in greater flexibility both regarding the user’s input and the agent’s output. The adaptation to the user relies on a complexity measure, allowing the agent to assess the level of complexity of the user’s input and thus to adapt the complexity level of its response. For the case of a simple travel information agent, the pattern extension procedure resulted in an increase in the number of available patterns of around a factor 20, greatly enhancing the agent’s capacity of adapting its output to the user.