Comprehensive Readability Assessment of Scientific Learning Resources

Readability is the measure of how easier a piece text is. assessment plays crucial role in facilitating content writers and proofreaders to receive guidance about easy or difficult In literature, classical readability, lexical measures, deep learning based model have been proposed assess readability. However, readability using machine data-intensive task, which requires reasonable-sized dataset for accurate assessment. While several datasets, indices (RI) models military agencies manuals, health documents, early educational materials, studies related computer science literature are limited. To address this gap, we contributed Computer (CS) AGREE , comprising 42,850 resources(LR). We assessed objects(LOs) pertaining domains Science (CS), (ML), software engineering (SE), natural language processing (NLP). LOs consists research papers, lecture notes Wikipedia topics list repositories CS, NLP, SE ML English Language. From statistically significant sample two annotators manually annotated LO’s difficulty established gold standard. Text was computed 14 Indices 12 measures (LM). RI were ensembled, used train The results indicate that extra tree classifier performs well on AGREE dataset, exhibiting high accuracy, F1 score, efficiency. observed there no consensus among shorter texts, but as length increases, accuracy improves. SELRD along with associated provide novel contribution field. They can be assessment, develop recommender systems, assist curriculum planning within domain Science. future, plan scale by adding more multimedia LOs. addition, would explore use methods improved