Introduction

Abstract

The main reason that current computer applications in symbolic processing fail to meet speed constraints on current machines is the gap between the applications and the languages and architectures in which they are implemented. Applications such as natural language understanding and symbolic equation solving, as compared with conventional applications such as numerical modeling and simulation, are further removed from conventional procedural/functional languages such as Pascal and Lisp and their corresponding numeric/scientific processor architectures. This is because these ambitious new applications must, in a sense, be written as meta-level interpreters. A meta-level interpreter is a program which performs additional levels of interpretation to implement features not present in the host language, e.g., nondeterminate execution for parsing or a reduction mechanism for theorem proving.