Homer: Black Box Knowledge Representation

Description

• Homer's architecture is modular with several modules each supporting their own distinct knowledge representations
• Some modules do not allow access to their knowledge by other modules despite being functionally linked to them
  • For example, the planner/reasoner, maintains its own episodic memory and representations of actions which the text generator is not able to access in order to explain its actions
• In addition not all modules were directly linked to each other, which lessened the degree to which modules could share knowledge