This paper describes a spoken dialogue system based on a generic specication of a cooperative communicating rational agent. We present some theoretical and practical apects of the overall approach, along with the speech-specic and natural language related issues raised by the eective implementation of the system. An account is also given of an evaluation of the system with naive users on the task of voice services directory (AGS) inquiry.

The development of natural multimodal dialogue systems remains a very difficult task. The flexibility and naturalness they offer result in an increased complexity that current software tools do not address appropriately. One challenging issue we address here is the generation of cooperative responses in an appropriate multimodal form, highlighting the intertwined relation of content and presentation. We identify a key component, the dialogic strategy component, as a mediator between the natural dialogue management and the multimodal presentation. This component selects the semantic information content to be presented according to various presentation constraints. Constraints include inherent characteristics of modalities, the availability of a modality as well as preferences of the user. Thus the cooperative behaviour of the system could be adapted as could its multimodal behaviour. In this paper, we present the dialogic strategy component and an associated platform to quickly develop output multimodal cooperative responses in order to explore different dialogic strategies.

A natural language generation system is typically constituted by two main components: a content planning component (e.g., text planner or dialogue act planner) and a linguistic realization component. But, this is not sufficient since, on the one hand, the message built by the content pldnning component is generally not adequately detailed in order to control the many possibilities for its expression and, on the other hand, the content planner cannot influence the way in which the message will be verbalized. Generation systems require a third component, called the micro-planning (or sentence planning or phrasing) component, which acts as an intermediary between the pragmatico-semantic level and the purely syntactic level. The micro-planner is responsible for transforming the message into a textual structure. For this transformation to be achieved, grammatical and lexical resources must be selected.

In the context of deployed spoken dialogue telecom services, we introduce a preprocessor called Fiction into the Spoken Language Understanding (SLU) component. It acts as an intermediate between the speech recognition and interpretation processes in order to increase the rate of utterances that are correctly rejected (CRR for Correctly Rejected Rate) without decreasing the rate of appropriately interpreted utterances. This component is based on statistical approaches of natural language treatment and contextual information. We also use active learning methods to determine the best training corpus size. On a deployed test corpus, the CRR increases from 60% to 86% and active learning method's results show that better performance can be achieved using fewer training data.

No abstract