The use of word-class distribution data for stylistics: Keats' sonnets and chicken soup

Systematic and objective descriptions of text's surface structures are necessary for analysis of style. One approach to syntax has been first to parse the text, then to count the proportions of word-classes. Simply computing the proportions (or percentages) does not allow valid comparison among texts, because the large proportion of one class will reduce the others. Furthermore, the classes are not equally likely to occur. It has been possible to select thirteen basic measures that assume linguistic relations among classes, e.g., the ratio of adjectives to nouns or auxiliary to main verbs, or that identify options, e.g., the distribution of nominal phrases between those with nouns or pronouns as heads. The values for the basic measures are considered to be statistically and linguistically independent, and ordinary tests can measure the significance of differences between texts. The computer algorithm which does the parsing cross-classifies words by their category (noun, adjective, auxiliary verb, etc.) and function (subject, complement, predicate, etc.). A style can be further defined by whether the constituents of nominal phrases with various functions are the same or not. The procedure is illustrated by data from Keats' sonnets and odes, Blake's Songs, and a data base representing other literary texts.