This thesis aims to create a proactive Intelligent Personal Assistant (IPA) within a car environment, utilizing location prediction and a face recognition to provide the context for that proactive behavior. A Dialogue Manager was implemented to provide a dialogue driven interface. This was realized as a multimodal interface using both voice and graphical input/output. The location prediction was realized with different neural networks to varying degrees of success. The project resulted in a prototype which acts like a proof of concept for a proactive IPA within a car environment, the prototype demonstrated the proactive behavior that the thesis aimed for. However most of the modules that make up the prototype, excluding face recognition, do not preform well enough to be considered complete, but they do demonstrate a proof of concept.
Johan Ekdahl and William Axhav Bratt