Conversational interfaces for task-oriented spoken dialogues: design aspects influencing interaction quality

摘要

This dissertation focuses on the design and evaluation of speech-based conversational interfaces for task-oriented dialogues. Conversational interfaces are software programs enabling interaction with computer devices through natural language dialogue. Even though processing conversational speech is a challenging problem, mainly because users’ spoken language can be extremely variable, the speech modality remains an attractive option because of its naturalness: speech is learned since childhood, that means, users neither need to learn nor to adapt to the designer’s interaction style. Also, speech can be very useful in situations when users cannot use other input modalities (e.g. while driving, accessing the interfaces over the phone, using pocket size devices or when impaired). The technological growth of the past decades brought conversational interfaces to a level of maturity which allows widespread application. Examples include interactive information systems, in-car applications, smart environments, media guides, training and educational systems, social robots, and so on. Therefore, the design and evaluation of conversational interfaces towards achieving a better interaction quality are as crucial as ever.