Effects of text essay quality on readers' working memory by a computational model

Assessment of essay quality, also called essay scoring, is a task that has been always carried out by human graders. Graders are usually asked to give their scores according to several determined linguistic/semantic criteria. These criteria are related to lexical, syntactical, semantical and discourse features of the texts. In order to replace human graders, automated essay scoring systems make use of statistics on the latter features in order to quantify the quality of the essays. However, there is a subjective component within the evaluation of the text quality that cannot be measured by artificial scorers. Text essays are a form of natural language communication and therefore they cause effects on readers and their cognitive functions. In the work presented in this paper, the dynamic effects that a read text causes on the working memory of readers are studied by means of a connectionist model of memory during reading. Besides, the correlation of those effects with the essay quality scores and text linguistic features is also analyzed. The biologically inspired model of memory includes mechanisms for emulating bounded cognition, getting a little closer to the BICA Challenge achievement. The results obtained also prove how BICA models can feedback Neuroscience and Psychology, thus closing the interdisciplinary loop.