Key feature analysis: a simple, yet powerful method for comparing text varieties

To date, corpus-based methods for comparing language varieties have fallen into one of two camps: ( 1) md analysis – a complicated multi-variate approach based on analysis of functionally motivated linguistic features in each text of a corpus, or ( 2) keyword/key pos analysis – simple, univariate techniques to identify any feature with a statistically skewed distribution in a corpus. In this paper, we introduce a complementary technique – key feature analysis – which is a simple quantitative approach to compare the texts in two varieties with respect to a set of functionally motivated lexico-grammatical features. We introduce the methods of key feature analysis, contrast them with other approaches for comparing text varieties, and present case studies from the domains of online registers and US presidential debates.