Natural Language Processing

NLP is among the oldest Artificial Intelligence efforts. The famous Turing Test was largely an NLP problem.

• Original applications were connected to machine translation of online texts (of Russian physics papers).
  • The original approach was word-by-word using dictionary lookup. This didn't work because of word-sense ambiguity. In one famous example, this:

    "In recent times Boolean algebra has been successfully employed in the analysis of relay networks of the series-parallel type."

    translated to this:

    "(In, At, Into, To, For, On, N) (last, latter, new, latest, lowest, worst) (time, tense) for analysis (and, N) synthesis relay-contact electrical (circuit, diagram, scheme) parallel- (series, successive, consecutive, consistent) (connection, junction, combination) (with, from) (success, luck) (to be utilize, to be take advantage of) apparatus Boolean algebra."

• The next approach was through grammatical analysis.
  • This started with the mathematical study of language in terms of their expressive power.
  • Noam Chomsky (1957) created the first taxonomy of languages: regular, context-free, context-sensitive, and unrestricted. With the
  • With the corresponding taxonomy of machines: finite automata, stack machine, bounded automata, and Turing Machine.

• This approach attempts to "understand" language from a structural perspective. Tokens are identified, parts of speech are assigned, and grammatical relations are also assigned: subject, verb, direct object, indirect object, prepositional phrase attachment, and so forth.

• This approach is prone to fail in the event of structural ambiguity, for example:
  • "The man hit the boy with the stick"
  • Who is holding the stick?

• Applications of NLP grew large and varied:
  • Natural Language Interface
  • Question Answering Systems
  • Text Processing (spelling and grammar checking)
  • Speech Processing
  • Machine Translation