Workpackage 3 Learning of Social Interactions​

Challenges

1. ​Learning interactions based on observation of humans.
   • Raw social interaction data is high-dimensional and noisy, which makes learning quite challenging​

2. ​Connecting interaction learning to a knowledge base.
   • To accomplish the long-term goal of dynamic real-world interaction, the system will need to generalize its behavior beyond the specifics that it has learned from the example interactions