Why might you want a computer to understand natural language?



Why might you not want a computer to understand natural
language?



What would it mean for a computer to really understand
natural language?



What components might contribute to the implementation
of a natural language understanding system?



How hard is this problem?

	The spirit is strong, but the flesh is weak.
	-> The vodka is good but the meat is rotten.



AI completeness





Levels of language analysis


Signal level (raw acoustical signal)

Phonetic level (language-specific sounds)

Lexical level (words and morphology)

Syntactic level (grammatical sentences)

Semantic level (meanings)

Pragmatic level (relations to goals, intentions, dialog roles, etc.)



In what order should these levels be processed?




Difficulties in computing natural language syntax


Language classes and the difficulty/possibility of
generation/parsing



Type 3 languages (regular).
Each production has only a single nonterminal symbol on its 
left-hand side and either a single terminal symbol or a 
terminal symbol followed by a nonterminal symbol on its 
right-hand side. Can be parsed by a finite state automaton.

Type 2 languages (context-free). The left-hand side of each 
production must always consist of exactly one nonterminal 
symbol. Can be parsed by a push-down automaton.

Type 1 languages (context-sensitive). The left-hand side of 
each production must not have a length greater than that of 
the right-hand side. Can be parsed by a linear bounded 
automaton.

Type 0 languages (recursively-enumerable). No restrictions. 
Can be parsed with a Turing machine.



Linguistic theories often make use of Type 0 languages, but 
because of their computational properties, most implemented 
systems are based on Type 2 languages.




Difficulties in computing natural language semantics



Assumptions about the audience made by speakers

  Representing events

  Representing complex human activities

  Semantic primitives

  Semantics of quantity

  Semantics of time

  Semantics of space

  Semantics of knowing and belief