Recently,a number of authorshave proposedtreating dialogue systems as Markov decision processes(MDPs). However,the practicalapplicationofMDP algorithms to dialogue systems faces a numberof severe technicalchallenges.We have built a general software tool (RLDS, for ReinforcementLearning for Dialogue Systems) on the MDP framework, and have applied it to dialogue corpora gatheredbased from two dialoguesystemsbuilt at AT&T Labs. Our experimentsdemonstratethat RLDS holds promise as a tool for "browsing" and understandingcorrelationsin complex, temporallydependentdialogue corpora.

This paper describes a novel method by which a spoken dialogue system can learn to choose an optimal dialogue strategy from its experience interacting with human users. The method is based on a combination of reinforcement learning and performance modeling of spoken dialogue systems. The reinforcement learning component applies Q-learning (Watkins, 1989), while the performance modeling component applies the PARADISE evaluation framework (Walker et al., 1997) to learn the performance function (reward) used in reinforcement learning. We illustrate the method with a spoken dialogue system named ELVIS (EmaiL Voice Interactive System), that supports access to email over the phone. We conduct a set of experiments for training an optimal dialogue strategy on a corpus of 219 dialogues in which human users interact with ELVIS over the phone. We then test that strategy on a corpus of 18 dialogues. We show that ELVIS can learn to optimize its strategy selection for agent initiative, for reading messages, and for summarizing email folders.

Cue phrases may be used in a discourse sense to explicitly signal discourse structure, but also in a sentential sense to convey semantic rather than structural information. Correctly classifying cue phrases as discourse or sentential is critical in natural language processing systems that exploit discourse structure, e.g., for performing tasks such as anaphora resolution and plan recognition. This paper explores the use of machine learning for classifying cue phrases as discourse or sentential. Two machine learning programs (Cgrendel and C4.5) are used to induce classification models from sets of pre-classified cue phrases and their features in text and speech. Machine learning is shown to be an effective technique for not only automating the generation of classification models, but also for improving upon previous results. When compared to manually derived classification models already in the literature, the learned models often perform with higher accuracy and contain new linguistic insights into the data. In addition, the ability to automatically construct classification models makes it easier to comparatively analyze the utility of alternative feature representations of the data. Finally, the ease of retraining makes the learning approach more scalable and flexible than manual methods.

GUS is the first of a series of experimental computer systems that we intend to construct as part of a program of research on language understanding. In large measure, these systems will fill the role of periodic progress reports, summarizing what we have learned, assessing the mutual coherence of the various lines of investigation we have been following, and suggesting where more emphasis is needed in future work. GUS (Genial Understander System) is intended to engage a sympathetic and highly cooperative human in an English dialog, directed towards a specific goal within a very restricted domain of discourse. As a starting point, GUS was restricted to the role of a travel agent in a conversation with a client who wants to make a simple return trip to a single city in California. There is good reason for restricting the domain of discourse for a computer system which is to engage in an English dialog.

In R. O. Freedle and J. B. Carroll (Eds.), Language comprehension and the acquisition of knowledge. Washington, D.C.: Winston, 41-69