Jump to the Overview of Homer page.
Jump to Homer: Architectural Details page.
Jump to the Home page.

The Text Interpreter

Homer's text interpreter takes English sentences as input then tranlates these into state transition representations . The interpreter's output goes into the episodic memory . Sentences can be statements of fact, commands, or questions.

The vocabulary is intended to comprise "about 1900 words, which are approximately the union of Ogden's Basic English (Ogden 1934) and the 1000 most frequent English words."[1] The Basic English vocabulary provides the breadth needed to describe objects and events in the environment. The frequent English fascilitates better communication with the user.

Homer can parse fairly complex sentences. The syntax covered is outlined below:

  1. verb type: intransitive, transitive, and bitransitive
  2. noun type: possessive, count and noncount
  3. adjective type: comparative, superlative
  4. tense: present, past, perfect, progressive, perfect progressive, and infinitive
  5. active and passive voice
  6. negation and conjunction
  7. nested prepositional phrases and relative clauses
( see sample dialogs for examples .)

Parsing is done bottom-up with semantic interpretation interleaved at each step. The result of parsing and semantic interpretation is an inform event. If the input is a command, the inform event takes the form of a goal for Homer to achieve. Similarly, a question is stored as an inform event with the goal of answering the question. The interpreter puts the inform events into the episodic memory.