The Comparative Power of Type/Token and Hapax legomena/Type Ratios: A Corpus-based Study of Authorial Differentiation

Abstract

This paper presents an attempt to verify the comparative power of two statistical features: Type/Token, and Hapax legomena/Token ratios (henceforth TTR and HTR). A corpus of ten novels is compiled. Then sixteen samples (each is 5,000 tokens in length) are taken randomly out of these novels as representative blocks. The researchers observe the way TTR and HTR behave in discriminating four novelists: Joyce, Woolf, Faulkner and Hemingway. When compared to the traditional statistical features (e.g. word length average, Sentence length average, etc.), TTR and HTR are by far more competent in comparing the distinctive quantitative behavior of each novelist. It turns out that TTR and HTR contribute more or less in creating a sort of statistical identity which can be used in giving a vivid comparison and discrimination of the four novelists involved in this paper. Nevertheless, HTR sounds more viable in achieving the discriminating task than TTR.