A way out of dead end situations in dialogue systems for human-robot interaction

In this paper, we present a strategy for resolving difficult situations in human-robot dialogues where the user input is inconsistent with the current discourse. Reasons for the inconsistency are analyzed in detail and a set of rules is implemented to take all of them into account. In a user test, we evaluated the success of the strategy which can reduce the communication problems resulting from misrecognized user utterances in human-robot communication.