Reshaping Automated Writing Evaluator for English as Foreign Language Users

To understand the role of the human rating and automatic writing evaluator (AWE), we compared their scores. There was a notable score difference in Focus and Purposes as a metric. This result indicated the need to validate and evaluate the metrics. Regarding AWE, more data is needed to improve the inference performance. Another improvement needs to include word embedding by pre-training the evaluator with a bigger corpus to represent better results.