Automatic Essay Scoring Model Based on Two-Layer Bi-directional Long-Short Term Memory Network

Automatic essay scoring provides a cost-effective and consistent alternative to human correction. However, in order to obtain good performance, human experts are needed to extract features of text manually in traditional ways. We propose a two-layer bi-directional long-short term memory model that a fully automatic essay scoring model. The agreement of essay scores which marked by our model and human raters respectively has achieved to 0.870 based on metrics of Cohen's κ. This model can achieve excellent results like human beings' professional raters.