The application of deep learning in automated essay evaluation

Abstract

The shift from Automated Essay Scoring (AES) to Automated Essay Evaluation (AEE) indicates that natural language processing (NLP) researchers respond positively to the request from language teaching field. Writers and teachers need more feedback about writing content and language use from AEE software beside a precise evaluative score. This requirement can be met by the neural network based deep learning technique. Deep learning has been applied in many NLP fields and great success has been made, such as machine translation, emotional analysis, question answering, and automatic summarization. Neural network based deep learning is suitable for AES research and development since AES requires mainly a precise score of writing quality. This can be accomplished with human accurately scored essays as input and scoring model as output with deep learning technology. However, AEE requires more than a score and deep learning can be used to select linguistically meaningful features for writing quality and apply in the AEE model construction. Related experiments already show the feasibility and further research is worth exploring.