The aim of this thesis is the development and testing of domain-independent algorithms and tools for predicting the dialogue outcomes in the framework of a spoken language dialog system. The application, in which the algorithms and methods will be demonstrated and evaluated, will be that of an automated agent providing customer care for technical problems, e.g. with cable TV and broadband internet. The envisioned methods are stochastically based and require huge amounts of log scripts as well as utterance transcripts representing interactions between the system and the user.
The thesis will be carried out under the auspices of the Graduate School Mathematical Analysis of Evolution, Information and Complexity and in cooperation with SpeechCycle, NYC, USA.