A unification of Halstead's software science counting rules for programs and english text, and a claim space approach to extensions

Abstract

In his Elements of Software Science, Maurice Halstead proposed that software quality measurements could be based on static lexemic analysis of the vocabularies of operators and operands, and the number of occurrences of each class, in computer programs. He also proposed that quality issues in Natural Language text could be addressed from similar perspectives, although his rules for programs and for English seem to conflict. This paper suggests that Halstead's seemingly disparate rules for classifying the tokens of programs and the tokens of English can be generally reconciled, although Halstead himself does not claim such a union. The thesis of Part One is a unification of his two procedures, based on a linguistic partitioning between "open" and "closed" classes. This unification may provide new inputs to some open issues concerning coding, and suggest, on the basis of a conceptual rationale, an explanation as to why programs which are by Halstead's definition "impure" might indeed be confusing to the human reader. Part Two of this paper, by exploring the nodes in a textual "Claim Space," briefly considers other groupings of the classes taken as primitive by Halstead, in ways which bring to light alternate and supplementary sets of candidate coding rules productive for study of textual quality.