Automated User-Centred Task Selection and Input Modification in Language Learning

Abstract

This article presents the results of a CALL-experiment in which reading material is selected dynamically, based on the “fit” between the vocabulary proficiency of individual students and the relative difficulty of texts. The texts were analysed, selected and presented online, together with a personalized electronic dictionary with words that were assumed to be unknown. In a pre-test – treatment – post-test design in which 32 Dutch L2-students took part, the vocabulary learned implicitly while reading the texts was measured. The relation between user-initiated noticing and word retention was also examined. We found an average word-learning improvement of 10.8%. On the basis of the non-significant differences between various proficiency groups, we concluded that the method we propose for the automated adaptive selection of reading texts ensures that learners of different proficiency levels receive linguistic input that is best fitted to their abilities. Using frequency information for both automated analysis of texts and the compilation of a personalized dictionary has great potential for more user-centred task selection and guidance. We also found a relation between user-initiated dictionary use and word retention, which was strong for the lowest proficiency level. With respect to the words looked up in the dictionary, a strong correlation was found between noticing and retention.