Article "Design and evaluation of automated writing evaluation modals: Relationships with writing in naturalistic settings"

Ayu Rina Dwi Utari
16611073
28th article

"Design and evaluation of automated writing evaluation modals: Relationships with writing in naturalistic settings"

This article describes the Automated Writing Evaluation (AWE) system built by extracting features from a 30-minute essay and using statistical capital to predict optimum human scores on a 30-minute essay. AWE's goal is to predict performance in realistic natural-world tigas. Therefore, the way to create feature weights in the AWE model is to choose optimized weights to predict real-world criteria. This unique new approach is used in a sample of 194 graduate students who provide two examples of their writing from the required graduate school paper. Contrary to the results of previous studies that predicted portfolio values, experimental models were no more effective than traditional models in predicting scores on actual writing performed from postgraduate.