elligent, could not give honest answers to personal questions such as "who were your parents.") Clearly, a fundamental precondition to passing this test was that the machine be able to interpret and respond in natural language.
Two more practical motives also played a part in the quest for natural-language processing. One was the problem of giving computers their operating instructions, a task that would be much easier if the machine could "understand" ordinary language, as opposed to rigid and specialized computer languages such as BASIC. The other was the desire to develop a computer or program which could translate between different languages--specifically, in the Cold War years of the 1950s and 1960s, to translate Russian-language scientific and technical journals into English. Each of these goals was interconnected. The linguist Noam Chomsky, for example, was closely involved in efforts to solve the problem of natural-language processing; the theories of grammar that grew out of his work influenced both our understanding of how the grammars of natural languages work and the development of more sophisticated computer-programming languages.
As we shall see, natural-language processing has proven to be an exceptionally difficult challenge; after some four decades of work, we are still unable to devise a computer program that can reliably translate natural language instructions into computer instructions; still less can any computer translate freely between natural languages, or pass a Turing test. Nevertheless, limited applications of natural-language processing have been developed which have been both interesting in the insights they give into human interactions, and useful in providing computers the capability to respond to natural-language commands in a wide variety of practical situations.
It is useful to approach the problem of natural-language processing from two directions: from machine instructions, by whi...