Automatic speech understanding systems are beginning to attain a Automatic speech understanding systems are beginning to attain a level of sophistication where commercial level of sophistication where commercial applications are within reach. However, if humans and machines a applications are within reach. However, if humans and machines are ever going to communicate in a natural way, it re ever going to communicate in a natural way, it is of vital importance that language modelling go beyond

more »

... e sen is of vital importance that language modelling go beyond the sentence level. A profound understanding of discourse tence level. A profound understanding of discourse structure is required, and to this end, knowledge concerning how structure is required, and to this end, knowledge concerning how prosody interacts with other linguistic phenomena prosody interacts with other linguistic phenomena is needed. Not only will better prosodic modelling of discourse is needed. Not only will better prosodic modelling of discourse lead to better speech recognition/understanding, it lead to better speech recognition/understanding, it will also yield more natural will also yield more natural--sounding speech synthesis. This paper reports on a dialogue/pros sounding speech synthesis. This paper reports on a dialogue/prosody project at Telia ody project at Telia Research, Sweden. A Wizard Research, Sweden. A Wizard--of of--Oz simulation of a computerized reservation system was used to c Oz simulation of a computerized reservation system was used to collect realistic ollect realistic speech data [pp. 2 speech data [pp. 2----3]. Fifty subjects were given three tasks each that entailed th 3]. Fifty subjects were given three tasks each that entailed the reservation of flights, trains, car hire e reservation of flights, trains, car hire and hotel reservations. To avoid linguistic influence on the sub and hotel reservations. To avoid linguistic influence on the subjects' utterances, the tasks were given as maps and jects' utterances, the tasks were given as maps and icons. A icons. A ToBI ToBI--style analysis was applied [p. 4], adapted to meet language style analysis was applied [p. 4], adapted to meet language--specific requirements [pp. 5 specific requirements [pp. 5----6]. The 6]. The dialogues were dialogues were analyzed analyzed with regard to phrase boundaries, tones, with regard to phrase boundaries, tones, disfluencies disfluencies, syntax (functions/categories), new vs. , syntax (functions/categories), new vs. given information and pitch range. This paper describes our obse given information and pitch range. This paper describes our observations concerning the interaction between rvations concerning the interaction between prosodic, syntactic and higher prosodic, syntactic and higher--level linguistic phenomena, such as discourse structure [ level linguistic phenomena, such as discourse structure [OHs OHs 9, 10, 11]. 9, 10, 11]. 1