Resolving paraphrases to support modeling language perception in an intelligent agent
Sergei Nirenburg, Marjorie J. McShane, Stephen Beale · 2008
When interacting with humans, intelligent agents must be able not only to understand natural language inputs but also to remember them and link their content with the contents of their memory of event and object instances.As inputs can come in a variety of forms, linking to memory can be successful only when paraphrasing relations are established between the meaning of new input and the content of the agent's memory.This paper discusses a variety of types of paraphrases relevant to this task and describes the way we implement this capability in a virtual patient application.