Natural Language Understanding in Homer

Natural Language Understanding in Homer

Homer is capable of both interpreting and generating natural language text. All communication with users is through English text. Homer's vocabulary is intended to encompass approximately 1900 words.

The text interpreter takes sentences as input then stores information from the sentences in the episodic memory. Sentences can be commands, questions, or statements of fact.

The text generator constructs sentences to convey several different kinds of information. First, Homer informs the user of its current actions and completed tasks. Second, it acknowledges all user commands and statements of fact. Third, Homer informs the user of realizations it has come to through reflection. Homer's use of language is very passive. The text generator does not relay information about Homer's knowledege or his surroundings unless the user asks specific questions.


Return to the top of this architecture.

Go to a discussion of this capability for multiple architectures.


Current Location: Homer-Capabilities-Natural Language Understanding

Go to NEXT page.