An analysis of navigational dialogues aboard a car

The study reported in this article was designed to contribute to the develo pment of on-board navigational aid systems. In order to improve the functio nal value of such systems, it is crucial to have information on how people actually exchange information in navigational tasks. With this purpose in m ind, we collected a corpus of navigational dialogues in realistic situation s involving a driver and a co-driver in a car in an urban environment. Twen ty-seven couples (driver plus co-driver) were assigned to one of four condi tions resulting from the combination of two factors. The first factor was w hether or not the co-drivers had any prior knowledge of the mute to be foll owed. The second factor was related to the mode of expression used by co-dr ivers (in particular, whether they had no constraints or were required to c ommunicate only verbally without gestures). The analysis of the data provid ed evidence that co-drivers' prior Knowledge influenced the content of dial ogues, in particular, the selection of landmarks and the expression of orie nting instructions. The dialogues involving co-drivers without any prior kn owledge contained more sentences, aid each sentence was longer than those i n dialogues involving co-drivers with prior knowledge. Constraints on the m ode of expression used by cc-drivers also moderately affected the dialogues . A detailed classification of the drivers' and co-drivers' utterances was established. Drivers' utterances were classified into sh categories: confir mations; queries; requests for repetitions or precisions; requests for conf irmation; reformulations; information to the co-driver. Co-drivers' utteran ces included six types: marking; descriptions of actions; descriptions of s cenes; advance descriptions; confirmations; requests for information. The s tudy also analyzed the descriptive strategies used by co-drivers when appro aching and traversing complex spaces, such as intersections. Three distinct patterns of dialogues: were described, depending on whether they occurred at a distant point (before the section in which the intersection would be f ound), in a nearby area (in which the intersection was visible) or in an im mediate area (where the vehicle had reached the intersection). The results of this study are discussed for their implications on the specification of navigational aid systems.