A Cycling Cancellation-Automaton for Sentence Well-Formedness

This paper attempts to specify what is the simplest device sufficient to recognize sentence structure, i. e. to state for arbitrary sequences of words whether they are well-formed sentences of a natural language. If each word in the sentence is represented by its explicit string-analysis relations, then the sufficient device is one which scans this string-representation of the sentence, cancels every sequence which consists of a symbol followed by its inverse, and repeats the scanning until no more cancellations occur. The device thus requires a memory only one symbol long. In addition, it must retain the uncancelled residue of the sentence after each scan, in order to cycle through it on the next scan; and storage is also needed for the various string-representations which a given sentence can have.

[1]  P A Shapiro ACORN--an automated coder of report narrative. , 1967, Methods of information in medicine.