SN AP - An experiment in natural

Computers are being used to a rapidly increasing extent, to manipulate and to generate materially, mechanically. (1) Many applications simply require items of information to be selected from a file of fixed format, heavily abbreviated records, and expanded into statements that are self-explanatory, and used perhaps in individual communications or incorporated in computer typeset compendia. At the other end of the spectrum are the interrelated challenges of mechanical indexing. abstracting, and translation. The burgeoning applications of computers to publishing, education, library work and information services in most major branches of science and scholarhsip are leading to a host of text processing and generating problems that span these limits of complexity. Many of the programming problems of mechanical text processing also arise in other kinds of symbol man 1 pulation, such as the compilation of computer programs, and the simplification of mathematical expressions. Several programming languages, such as COMIT and SNOBOL have b,een developed over the years for mechanical symbol manipUlation, and used to process text. FORTRAN, supplemented by some simple assembly language subroutines, has been applied to non-numeric problems quite extensively. ALGOL and COBOL also have been used this way. Several features of PL/1 will facilitate its use in processing text. A number of languages and processors have been developed for special kinds of text processing problems, such as mechanical editing. SNAP is a procedural language for nonscientists to use. A SNAP procedure consists of a sequence of statements that have the appearance of simple English sentences. The primary rationale of SNAP is that mechanical text processing requires a language that many nonscientists will be willing to learn, and which they will find easy to use. Since such people deal primarily with English sentences in their daily work, a programming language that is a stylized subset of English should seem to them more "natural" than one that is symbolic. This rationale is not negated by the fact that nonscientists are using symbolic languages effectively. Many scientists once learned to write in assembly language, but the number of people programming scientific problems was greatly increased by FORTRAN. Several SNAP constructions are quite like COBOL. SNAP is designed like BASIC, to enable a beginner to get useful results after learning very little; and to proceed by incremental learning efforts to deal with problems of increasing complexity. In its primary emphasis on string handling, however, SNAP is different; and in the details of the instruction set, and how the basic instructions are expressed, and in the proposed use of statements that invoke subroutines, to allow multitudinous language extensions. SNAP has some novel features. A prototype processor for the basic SNAP language, that is SNAP without the subroutine capability, was implemented in a compatible subset of FORTRAN IV and now runs on several models of computers.