The ability to communicate in natural language constitutes an absolute prerequisite for integrating robots in safety-critical environments. Current approaches that focus on an agent learning through interaction are increasingly relying on end-to-end training. A major challenge in this direction is how an agent can overcome ambiguity phenomena commonly observed in human conversation. The proposed project aims to investigate neural approaches in conversational Artificial Intelligence, which can benefit from repair mechanisms through clarification dialogue in a collaborative setting towards more accountable and transparent Human Robot Interaction.
I am a first year PhD student in the CDT of Robotics and Autonomous Systems at Heriot-Watt University. My research interests are within transparent multimodal agents for human-robot collaboration. Prior to that, I completed an MSc in Artificial Intelligence with Speech and Multimodal Interaction at Heriot-Watt University. My past experience also includes developing systems as a Machine Learning Engineer for Speech Emotion Recognition.