Computer Modelling of English Grammar

Recent work in artificial intelligence has developed a number of techniques which are particularly appropriate for constructing a model of the process of understanding English sentences. These methods are used here in the definition of a framework for linguistic description, called "computational grammar". This framework is employed to explore the - details of the operations involved in transforming an representation English sentence into a general semantic Computational grammar includes both "syntactic" and "semantic" constructs, in order to clarify the interactions between all the various kinds of information, and treats the sentence-analysis process as having a semantic goal which may require syntactic means to achieve it. The sentence-analyser is based on the concept of an "augmented transition network grammar", modified to minimise unwanted top-down processing and unnecessary era bedding. The analyser does not build a purely syntactic ,structure for a sentence, but the semantic rules operate hierarchically in a way which reflects the traditional tree structure. The processing operations are simplified by using temporary storage to postpone premature decisions or to conflate different options. The computational grammar framework has been applied to a few areas of English, including relative clauses, referring expressions, verb phrases and tense. A computer program ( "MCHINE") has been written which implements the constructs of computational grammar and some of the linguistic descriptions of English. A number of sentences have been successfully processed by the program, which can carry on a simple. dialogue as well as building semantic representations for isolated sentences