How to Access Large Navigation Databases in Cars by Speech

Authors: André Berton, Sandra Mann, Peter Regel-Brietzmann

Abstract:

Navigation applications are becoming increasingly complex, since databases and features are rapidly growing. Navigation databases include a growing number of points of interest (POI) and a more precise resolution at the street, crossroad and house number level. Human-Machine Interfaces (HMI) such as Speech Dialog Systems (SDS) need to be designed in order to allow comfortable access to data with as little driver distraction as possible. We discuss how the user can find the intended data in such a large database? Current approaches require browsing the hierarchical database structure through categories and sub-categories, such as restaurant and vegetarian. This paper presents a new interface that utilizes a contextfree keyword search, and relies on a hierarchy of categories only for disambiguation purposes. This allows a user who does not completely know the hierarchy to find a destination by simply saying it. The user does not even have to know the precise name of the destination, since the proposed method generates wording variants by decomposing the name and recombining its parts. After presenting the new interaction design and dialog requirements, we derive a dialog system architecture which we implemented to fulfill these requirements. A short pre-evaluation indicates that users prefer the new interaction method to the interaction of the S class series model, because of its increased comfort, ease of use, general appeal, and in particular because of its voice control interface.