A Cognitive Pattern for Comprehension and Recall of Narrative and Expository Texts: An Interdisciplinary Approach to Memory Modeling in Natural Language Processing Systems

In recent years, cognitive modeling of textual memory has gained significant attention at the intersection of cognitive science and artificial intelligence. However, most existing language models lack alignment with the actual cognitive processes involved in human comprehension and recall. This article introduces and analyzes a hybrid framework—referred to as the Mirroring Model—designed to more accurately represent the cognitive mechanisms underlying the understanding and retention of narrative and expository texts. The model is structured around three core components: (1) event indexing based on a five-dimensional schema (time, space, agent, goal, and causality); (2) activation of conceptual lexical networks within textual context; and (3) positional weighting of sentences according to the serial position effect in memory. This integration results in a model that is more consistent with human memory in terms of narrative structure, semantic layering, and cognitive organization of information. In this study, the model was applied to three test texts and its outputs were compared with participants’ behavioral recall data. Four key evaluation metrics—accuracy, precision, recall, and F1-score—were employed to assess performance, and cognitive correlation between the model and human memory was examined using Pearson’s correlation coefficient and the t-test. Findings indicate that, compared to two widely-used models (the Landscape Model and the Event-Indexing Model), the Mirroring Model achieves higher cognitive correlation with actual recall data and provides greater accuracy and balance in predicting memory performance.