On the importance of root‐stem truncation in word‐frequency analysis