The development of natural-language processing, or the ability of computers to respond to commands or other inputs couched in ordinary English, has been one of the central challenges of computer science, and one closely related to the issue of artificial intelligence (Goshawke, Kelly, and Wigg, 1987). The general problem of natural-language processing has not yet been solved, but elements of natural-language processing have been employed in a highly useful category of applications known as expert systems.
When the first electronic digital computers were introduced in the late 1940s, their ability to swiftly perform elaborate sequences of calculations--for example, predict the trajectory and control the flight of a rocket--fascinated both specialists and the general public. Their computative abilities, in some respects superhuman, seemed suggestive of human intelligence as the capabilities of no previous machines had been, and early computers (though no faster or more powerful, in fact, than a present-day programmable pocket calculator) were sometimes referred to as "giant electronic brains."
By the 1950s, serious efforts were underway to adapt computers to the task of interpreting ordinary language, or as it was formally called, natural language. A variety of motives, theoretical and practical, lie behind these efforts. On the most theoretical level, natural-language processing was major feature, perhaps the central feature, of artificial intelligence (AI). The classical definition of AI was the so-called Turing test, which held that a machine must be regarded as intelligent if a person communicating, say, via keyboard, could not tell if the responses he or she received came from a man impersonating a woman or from a machine impersonating a man impersonating a woman (Schank, 1984, pp. 58-59). (The roundabout procedure was perhaps intended to introduce a necessary element of deception; after all, a computer, however int...