Text categories and where you can stick them : A crude formality index

This paper applies principal components analysis (PCA) to solve the problem of interpreting pre-existing corpus text categories for analysis of linguistic variation. The method is illustrated by constructing an index of the complex notion "formality " from PCA of a set of high-frequency wordform-based counts. The first principal component from this analysis acts as a broad formality index; a second principal component is tentatively identified as marking "concrete facts" versus "abstract discussion"'. Subsequently, text categories from the corpora are positioned on these textual dimensions, and selected categories are evaluated for internal consistency by comparing the distribution of texts across subcategories. Finally, suggestions are made concerning further developments and applications of the method used here, and its implications for the use of corpora in variation studies.