Spoken multimodal human-computer interfaces constitute an emerging topic of interest not only to academia but also to industry. The ongoing migration of computing and information access from the desktop and telephone to mobile computing devices such as Personal Digital Assistants (PDAs), tablet PCs, and next generation mobile phones poses critical challenges for natural human-computer interaction. Spoken dialogue is a key factor in ensuring natural and user-friendly interaction with such devices which are meant for everybody. Speech is well-known to all of us and supports hands-free and eyes-free interaction, which is crucial, e.g. in cars where driver distraction by manually operated devices may be a significant problem. Local companies working in the field of spoken natural language dialogue systems will provide guest lectures.

1. Natural Language Understanding.

This part of the lecture covers morphological, syntactic, and semantic processing of spoken natural language from both a linguistic and an algorithmic perspective. Starting out with the traditional view of natural language as a formal language, the focus is shifted to modern quantitative techniques in natural language understanding: using large corpora, statistical models for acquisition, disambiguation, and parsing. Also, it examines and constructs representative systems.

2. Dialogue Modelling.

In this part of the lecture, the basic concepts and approaches in dialogue research will be introduced, and their practical usage in dialogue systems be reviewed. Fundamentals of human dialogue modelling, such as conversational participant's goals and intentions, dialogue structure and topic tracking, and reasoning about rationality and cooperation will be covered. Theoretical underpinnings in regard to practical design requirements will be discussed, and in particular, focus will be placed on on different styles of dialogue management (system directed, mixed initiative), and special requirements speech interfaces impose on dialogue management (disfluencies, misunderstandings, non-understanding). Practical dialogue modelling languages used today in commercial dialogue systems, such as VoiceXML and GDML, will also be discussed. Finally, an overview of dialogue evaluation methods will be given.

3. Exercises and Practicals.

Spoken natural language dialogue systems development with a focus on parsing and practical dialogue management.