Swiss Voice in Mycroft

Abstract

Voice assistants have become ubiquitous to the point that most people carry one in their pockets at all times, embedded into their smartphones. They are already being actively used by millions of users worldwide to get information about current weather, listen to the latest news or control their smart homes. Due to the unique nature of the Swiss German language of being only a spoken language, with German being the official language of the land, voice assistants typically do not speak nor understand Swiss German. This introduces an additional barrier to voice assistants’ widespread use in Switzerland’s German-speaking parts in the form of feeling unnatural to use, as communication does not occur in the user’s true mother tongue. This thesis examines the open-source Mycroft voice assistant and the suitability of its design for bridging the gap between voice assistants and Swiss German users, by showing how to integrate a custom-built text-to-speech engine into the Mycroft platform. This is accomplished by leveraging Mycroft’s available German language support as well as its adaptability and flexibility. Furthermore, the thesis shows how to adapt existing skills and develop new functionality for the adapted Mycroft platform.