Characterizing prior knowledge-attention relationship by a computational model of cognitive reading

Interest and prior knowledge are supposed to influence reading comprehension and learning from natural language texts. The effects of interest have been well studied in the literature, but little effort has been made on empirically establishing the influences of prior knowledge in reading attention and engagement, and therefore in comprehension and learning. A quantitative characterization of this relationship is proposed in this paper by means of a connectionist and computational method, a model of cognitive reading which allows to configure and isolate inferential depth and memory issues, which are well-known to be strongly related to attention and engagement. Results have pointed out a clear and straight relationship between prior knowledge and the latter issues and they have shown the computational model to be suitable as experimental framework for the validation of further hypothesis related to human language processing.