User initiated learning for adaptive interfaces

Abstract

Intelligent user interfaces employ machine learning to learn and adapt according to user peculiarities. In all these cases, the learning tasks are predefined and a machine-learning expert is involved in the development process. This significantly limits the potential utility of machine-learning since there is no way for a user to create new learning tasks for specific needs as they arise. We address this shortcoming by developing a framework for user-initiated learning (UIL), where the end user can define new learning tasks, after which the system automatically generates a learning component, without the intervention of an expert. We describe the knowledge representation and reasoning required to replace the expert, so as to automatically generate labeled training examples, select features, and learn the required concept. We present an implementation of this approach for a popular email client and give initial experimental results.