EYEBALL: A computer program for description of style

Abstract

P I • e EYEBALL information system is designed, for students of literature, with special applications to literary style, criticism, and the history of ideas. It is composed of a series of computer programs which surveys works of literature in natural (or real) language and, by two stages of intermediate processing, with the use of real-time hardware 1 when possible, "marks" many of the text's properties of syntax and vocabulary use. The major role of the intermediate processing is to indicate the ends of phrases (prepositional, infinitive, and so on) and of clauses and locations of subjects, predicates, and objects within clauses. There seems to be no way to perform these operations automatically because of the high degree of ambiguity in English ("love" can be either a noun, verb, or adjective). Such ambiguity causes continual conflict between the two goals of maximizing accuracy and minimizing the man-hours involved in the analysis. In general, we have attempted to design algorithms (computing procedures) which will function with an error frequency of the order of magnitude of some two percent. For individuals schooled in the humanities, such error tolerances may seem overwhelming, although scientists and social scientists will consider them quite acceptable. In Ross' experience in making hand counts, which involved going through a text two or more times, there were often errors up to ten percent, even after double checking. If completely accurate data are needed for certain parts of an analysis, options are available to check on all decisions made by the machine and to impose individual corrections.