LAUSR

This study introduces the application of deep-learning technologies in automatically generating guidance for independent reading. The study explores and demonstrates how to incorporate the latest advances in deep-learning-based natural language processing technologies in the three reading stages, namely, the pre-reading stage, the while-reading stage, and the post-reading stage. As a result, the novel design and implementation of a prototype system based on deep learning technologies are presented. This system includes connections to prior knowledge with knowledge graphs and summary-based question generation, the breakdown of complex sentences with text simplification, and the auto-grading of readers' writing regarding their comprehension of the reading materials. Experiments on word sense disambiguation, named entity recognition and question generation with real-world materials in the prototype system show that the selected deep learning models on these tasks obtain favorable results, but there are still errors to be overcome before their direct usage in real-world applications. Based on the experiment results and the reported performance of the deep learning models on reading-related tasks, the study reveals the challenges and limitations of deep learning technologies, such as inadequate performance, domain transfer issues, and low explain ability, for future improvement.