Automatically assessing children's writing skills based on age-supervised datasets

Abstract

In this paper, we propose an approach for predicting the age of the authors of narrative texts written by children between 6 and 13 years old. The features of the proposed model, which are lexical and syntactical (part of speech), were normalized to avoid that the model uses the length of the text as a predictor. In addition, the initial features were extended using n-grams representations and combined using machine learning techniques for regression (i.e. SMOreg). The proposed model was tested with collections of texts retrieved from Internet in Spanish, French and English, obtaining mean-absolute-error rates in the age-prediction task of 1.40, 1.20 and 1.72 years-old, respectively. Finally, we discuss the usefulness of this model to generate rankings of documents by written proficiency for each age. © 2014 Springer-Verlag Berlin Heidelberg.