Dialogue Systems Using Web-based Language Tools

Sammanfattning: Chatbots in commercial environments are on the rise with the release of several web-based language understanding tools. The vast majority of the dialogue systems deployed today uses very primitive state-machine architectures to model their interactions. These primitive approaches are reliable and easy to implement but the dialogue becomes very unnatural and the system always has the initiative in the conversation. The positive features of being easy to build, and the ability to easily retain control over the system normally supersedes the shortcomings.This thesis proposes a dialogue model that utilizes new approaches for dialogue modeling but aims to be easy to congure. The proposed dialogue management strategy is implemented in a prototype dialogue system. Developers are able to model their dialogues using an XML dialogue description. The system utilizes LUIS.ai, a recently launched web-based language tool for sentence analysis. LUIS.ai is evaluated together with the prototype dialogue system.i