No abstract

Confidence annotation allows a spoken dialog system to accurately assess the likelihood of misunderstanding at the utterance level and to avoid breakdowns in interaction. We describe experiments that assess the utility of features from the decoder, parser and dialog levels of processing. We also investigate the effectiveness of various classifiers, including Bayesian Networks, Neural Networks, SVMs, Decision Trees, AdaBoost and Naive Bayes, to combine this information into an utterancelevel confidence metric. We found that a combination of a subset of the features considered produced promising results with several of the classification algorithms considered, e.g., our Bayesian Network classifier produced a 45.7% relative reduction in confidence assessment error and a 29.6% reduction relative to a handcrafted rule.

No abstract

Opponent behavior in today's computer games is often the result of a static set of Artificial Intelligence (AI) behaviors or a fixed AI script. While this ensures that the behavior is reasonably intelligent, it also results in very predictable behavior. This can have an impact on the replayability of entertainment-based games and the educational value of training-based games. This paper proposes a move away from static, scripted AI by using a combination of deliberative and reactive planning. The deliberative planning (or Strategic AI) system creates a novel strategy for the AI opponent before each gaming session. The reactive planning (or Tactical AI) system executes this strategy in real-time and adapts to the player and the environment. These two systems, in conjunction with a future automated director module, form the Adaptive Opponent Architecture. This paper describes the architecture and the details of the deliberative and reactive planning components.

Abstract. The introduction of a coach competition in the RoboCup-2001 simulation league raised many questions concerning the development of a "coachable" team. This paper addresses the issues of dealing with conflicting advice and knowing when to listen to advice. An actionselection architecture is proposed to support the integration of advice into an agent's set of beliefs. The results from the coach competition are discussed and provide a basis for experiments. Results are provided to support the claim that the architecture is well-suited for such a task.