Action-oriented dialog modeling in a learning scenario

This project is based upon a social learning scenario with a humanoid robot. Not only for the immediate goal of learning but also for the interaction many interesting challenges arise.

• Social learning setting: interaction plays a crucial role
• Embodiment and situatedness:
  • Humans will utilize the entire body for communication rather than only voice
  • Humans will refer to objects and events in the world verbally as well as non-verbally

We follow that speech alone will not suffice for the interaction to be successful. Therefore we need to consider multimodal information and most especially actions within interaction:

• Moving or passing objects in order to establish shared attention
• Demonstrate functionality of objects or an action to learn
• Actions like approaching or leaving the communication partner Approach
• Considering actions as further modality and modeling their communicative function
• Grounding-based interaction modeling: assumption that communication partners share the goal to establish common understanding during a dialog
• Extending notion of grounding: actions as a presentation or an acceptance of a previous contribution
• User studies to prove our concept Summary

A fundamental axiom of linguistics is that communicating means acting. However, in the given scenario we focus on the reverse that acting also means communicating. Implementing our action-oriented dialog model on a real robot system we aim to realize a more intuitive interaction with the robot. [view:groupmembers==205]