The predictive capabilities of mathematical models for the type-token relationship in English language corpora

We investigate the predictive capability of mathematical models type-token relationship applied to vocabulary growth profiles selected English language documents. compare existing Good-Toulmin and Heaps formulae with an alternative approach based on Bernoulli trial word selection from a fixed finite using Zipf Zipf-Mandelbrot probability distributions. make two major observations: firstly, while model makes better predictions than model, optimized parameters latter correlate those former statistics gleaned independently data. Secondly, mean Zipf-Mandelbrot, provides more consistent unbiased prediction any individual alone.