Dialogue processing in spoken language systems. ECAI '96. Workshop, Budapest, Hungary, August 13, 1996. Revised papers (Q5906781)

This volume contains a selection of fifteen extended and revised versions of papers presented at the European Conference on Artificial Intelligence (ECAI'96) Workshop on Dialogue Processing in Spoken Language Systems. After an overview of the current state of the field, various technical issues are dealt with. These include user errors in spoken human-machine dialogue, different dimensions of a dialogue taxonomy, a Wizard-of-Oz study (hidden operator simulation) in the Verbmobil project, the classification of public transport information telephone dialogues using an information based coding scheme, a language generation perspective on speech production in human-machine dialogues with focus on the selection of appropriate intonation, input segmentation of spontaneous speech into semantic dialogue units in the speech-to-speech translation system Janus, an empirical study and transcript encoding scheme of pause units (segments bounded by natural pauses) in spontaneous Japanese dialogues, syntactic procedures for the detection and transformation of self-repairs in German dialogues, a discourse-level definition of utterance units in spoken dialogues, discourse structure principles underlying the ability of a computer system to conduct natural dialogues with humans, the architecture of Dragon Dialogue, a domain and application independent spoken language system, and the minimization of cumulative errors in the recognition of speech acts utilizing discourse context in the Enthusiast Spanish-to-English speech-to-speech translation system. The remaining papers focus on evaluation methodologies in terms of an automatic, system-to-system evaluation environment for spoken dialogue systems, an end-to-end evaluation method of translation performance on unseen data in the Janus speech-to-speech translation system, and a task-based evaluation methodology of the Trains-95 dialogue system.