This thesis is concerned with the development of a robust and efficient speech-enabled query interface for the travel information system. The approach taken is separated into three distinct processes: i) development of a directed-dialogue speech-enabled interface for a medium grammar based bus travel information system. This interface directs the user through a sequence of questions and answers to get the enquired result; ii) development of a multimodal interface that employs a mixed-initiative grammar to overcome the usability problems identified in the medium grammar directed-dialogue system. This interface allows the system to process a more natural language style of input rather than directing the user through a rigid sequence of questions and answers; iii) development of a directed-dialogue speech-enabled interface for an equivalent large grammar based bus travel information system that uses a novel method for real-time grammar segmentation and recognition. This thesis firstly presents the dialogue design and usability evaluation of a directed-dialogue speech-enabled query interface for a bus travel information system. The evaluation, based on a usability-engineering paradigm, analyses four human factors of the user interface: effectiveness, efficiency, user satisfaction and learnability. The initial interface design contributes a baseline specification for the construction of a speech-enabled interface and a usability test method for other speech application developers. This evaluation also highlights the usability issues associated with the use of directed-dialogue and speech-only interfaces. A mixed-initiative dialogue combined with a multimodal interface is then presented that successfully addresses all of the usability issues identified in the directed-dialogue interface. The good usability results reported for this improved interface show that the use of a mixed-initiative dialogue combined with a multimodal interface is an effective method for building a speech-enabled phone-based HCI system. Finally, this thesis introduces a novel last-word recognition based grammar segmentation method that is used to handle the large grammar issues associated with producing a real-time bus travel application. Large grammars tend to produce relatively slow recognition interfaces and this work shows how this limitation can be successfully addressed. This investigation therefore contributes a method for designing real-time speech-enabled interfaces that need to use very large grammars.