Evaluation of Lexical Diversity in Children’s Writing

Abstract

We examined the reliability of 24 different lexical diversity indices using children's compositions. It is desirable that the lexical diversity index is uncorrelated with the text length, but in the case of children's writing, uncorrelatedness does not ensure the reliability of the index because children's language ability is correlated with both the text length and the lexical diversity. Therefore, we measured indices cumulatively in the same composition, and evaluate the smallness of its variation by LMM. In addition, we attempted to modify three types of indices by using LMM to determine its parameter. It showed that the Simpson indices, the iterative measurement methods, and the LMM-corrected indices gave relatively good results. We also examined the relationship of these indices to the author's grade level and the teacher's evaluation of their writing, and it was confirmed that lexical diversity increased as the grade level increased and that the teacher's evaluation did not place much importance on lexical diversity.