Abstract
Having set up several dialogue systems for multiple domains we realized that we need to introduce each system thoroughly to the users in order to avoid misunderstanding. Even with well-described systems, users still tend to request out-of-domain information and are often confused by the system response.
In our ongoing work, we try to address these issues by allowing our conversational agent to speak about its abilities. Our agent is able to simply describe what it understands and why it decided to perform one of its summary actions.
In this short paper, we present our current system architecture.