Masterarbeit
User Preference Customization for the Smart Home
Completion
2023/07
Research Area
Students
Meshu Deb Nath
Advisers
Description
Nowadays, the usage of IoT devices by end users is increasing rapidly. IoT applications offer a versatile methods to provide end users the ability to control and monitor their smart home devices in a more flexible way to improve quality of life and convenience. In the last years the smart home device market has increased, according to the statistica, the IoT market will increase to about 182 Billion USD until 2025. Due to the increasing popularity, it is also important for the end user to have an easy way to configure and control their devices. The field of End User Development (EUD) focuses in particular on improving the interaction of non-technical end users with software-based systems. On the other hand, one popular and convenient way to control smart home devices is by using a voice interface like Amazon Alexa, Google Assistant or Apple Siri. These voice interfaces allows the end user to interact with their smart homes verbally without being bind to a graphical user interface. In this regard, goal-oriented solutions are a new paradigm that enable end users to specify their desires based on a high-level goal. A recent solution called GrOWTH is a goal-oriented voice-based approach that enables end users to easily program the diverse behaviours of their home using high-level goals. To achieve this, GrOWTH proposes VISH (Voice Interface for Smart Home), which is a natural language understanding and goal classification interface of user utterances (e.g., “I’m freaking freezy”). VISH deals with receiving a goal from the end user and automatically converts it to machine-understandable form. However, current solutions in the area of voice-based user interfaces for the smart home domain do not consider the preferences of the end user for goal specification. This means that the system does not allow the customization of certain behaviours. For instance, a goal like “I want to relax” has different meanings for different users. Some solutions which enable preference management use the concept of skills using visual-based method. However, voice-based solutions are still limited in allowing the end user to communicate their preferences via voice using a natural mechanism. Another important issue is the way failures are handled in voice-interfaces. Many solutions in dialogue management systems propose to either divert the user attention to features that the system can handle or allow the user to change their input. Therefore, the goal of this master thesis is to design a suitable approach or a combination of approaches to solve the outlined problems and provide a more natural and intuitive approach for receiving the end user preferences and handling failure in natural language understanding using voice-based methods. This includes in particular the examination of the state of the art regarding solution such as preference management in end user development solutions, preference handling via voice-based user interfaces and failure handling. The demonstration with smart home devices within the home, integration with the existing GrOWTH tool and empirical evaluation by conducting usability studies with end users and comparing with at least another voice-based interface (e.g., Amazon Alexa) with suitable metrics is also an important part of the work.
