|Title||Cognitive Models of Failure and Recovery in Natural Language Interactions: A Joint Actions Approach|
|Publication Type||Conference Paper|
|Year of Publication||2011|
|Authors||Murugesan, A, Frost, W, Brock, DP, Perzanowski, D|
|Conference Name||15th Workshop on the Semantics and Pragmatics of Dialogue|
|Conference Location||Los Angeles, CA|
Natural language interaction, like any other joint action, is a coordination problem involving agents who work together to convey and thus coordinate their interaction goals. Joint actions frequently fail, as agents act on their best guesses of what is intended by the other person. The ability of agents to correct each other, and recover from failures, makes it possible for joint actions to succeed even in highly error prone situations. In the modeling work presented here, a sequence of interrelated modules, originally developed in the Polyscheme cognitive architecture to understand simple commands to video application, is modified to implement error discovery and accommodate possible user-initiated repairs.
NRL Publication Release Number: