ABSTRACT
Using news articles for science assessment helps students and teachers increase their interest, motivation, and engagement in science learning. Students can understand how their knowledge from classroom instruction is linked to real-life scenarios from news articles. Despite the abundance of online science resources provided for students and teachers, finding and evaluating appropriate science learning resources is a daunting task. This is due, in part, to the lack of a systematic method for managing and evaluating science articles with up-to-date information that satisfy curriculum standards. The purpose of this study is to introduce a technology-enhanced framework based on machine learning and various natural language processing techniques to understand and evaluate science articles as they relate to specific curriculum standards. Our framework used Latent Dirichlet Allocation (LDA) to organize the topic structure from 1,025 recently published science articles for K to 12 students. Then, our system evaluated whether the articles contain science concepts that are closely related to curriculum standards. The best model identified an interpretable topic that can be used to accurately classify the science articles based on their curriculum standards. The system was also used to generate constructed-response test items using classified articles based on curriculum learning outcomes.
