Four A Procedural Model of Language Understanding

Much of the research on language is based on an attempt to separateit into distinctcomponents-components that can then be studied independently. Modern syn-tactic theoreticians have been tremendously successful at setting up complexruleswhich describe in detail the possible orderings of syntactic constituents- at thesame time other researchers are trying to define semantic relations and to modelthe cognitive structures underlying language use. Most attempts to model language understanding on the computer have followedthis strategy of dealing with a single component of language. They are constructedprimarily as a syntactic program (Kuno, 1965], a model of semantic connections(Schank, 1971), or an attempt to model the memory structures (Quillian 1967)Question-answering systems have had to deal with the entire language process but they have been severely limited in the breadth of their language ability Theonly attempt to handle large portions of language data was the machine translation effort, and it soon became obvious that the methods were not up to the requirementsof the task. Language translation could not be treated as a problem of rearrangingsyntactic structures and words, because attention to meaning was required even toachieve moderately acceptableresults. One basic limitation of those programs that have tried to handle the problems of meaning is that they have dealt almost exclusively with the understanding of singl