A new method of representing constituent structures.
The class of acceptable expressions (with no restrictions on length) of a natural language appears to form a regular set. The class of grammatical structures that corresponds to that class of expressions, however, does not, if those structures are represented in tree or bracketing form. We propose a parenthesis-free method of representing grammatical structures which permits the class that corresponds to the acceptable expressions of a natural language to constitute a regular set. The results of this method can be considered the basis of a viable model of human representation of grammatical structure for two reasons. First, the grammatical structures of acceptable expressions may be computed in tandem with the processing of those expressions from left to right; and, second, this computation can itself be carried out by a finite automaton. We apply this method to a fragment of English and show how it works for that fragment. A portion of this application has been implemented in PL/I.