Natural Language as an AI Problem

Natural language understanding is a classical AI Problem:

• Minimal Input Data: the natural language statement does not contain the message, but is a minimal specification to allow an intelligent reader to construct the message.

• Knowledge Based: the interpretation of the message is based in large part on the knowledge that the reader already has.

• Reference to Context: the message implicitly refers to a context, including what has been said previously.

• Local Ambiguity: many wrong interpretations are superficially consistent with the input.

• Global Constraints: there are many different kinds of constraints on interpretation of the input.

• Capturing the Infinite: a language understanding system must capture, in finite form, rules sufficient to understand a potentially infinite set of statements.