The larger the corpora we look at, the more word types we find. Assume $|V|$ is the number of types and $N$ is number of tokens. k and $\beta$ are positive numbers and $0< \beta <1$. Then$|V|= k N^{\beta}$