Paper Title
A Comparative Study of BI-LSTM and GRU for An Essay Auto Grading System

Abstract
In Today’s age, there are various examinations where students are required to express their thoughts in the form of essays. This gives rise to intelligent auto-grading systems that would alleviate the load on reviewers. This paper presents a comparative study of essay auto-grading system based on Deep Neural Network techniques bi-directional long short-term memory and gated recurrent unit combined with natural language processing. The proposed system encodes an essay as sequential embedding and harnesses a bi-directional LSTM and GRU models to catch the semantic information. It also draws attention to focus on sound information and captures specific features. The dataset used for training and testing is publicly available essay set Automated Student Assessment Prize on Kaggle. The results show that our system provides comprehensive evidence to differentiate the working of LSTM and GRU models. Also, the system focuses on significant words, their meaning, syntax and their relationships with the context by preserving semantic information to predict accurate results. Keywords - Long Short-term Memory, Gated Recurrent Unit, NLTK, Natural Language Processing, Neural Network.