Automated Essay Scoring System

Introduction

Automated Essay Scoring (AES) is a technology that uses computer algorithms to evaluate and score written essays. The goal of this project is to develop a robust AES system that can accurately assess essays based on various linguistic and structural features. This project proposal outlines a system that leverages machine learning techniques to improve the accuracy and efficiency of essay scoring.

Background

Recent research has highlighted the potential of machine learning models, particularly neural networks, in enhancing the performance of AES systems. These systems evaluate essays based on dimensions such as topic relevance, organization, coherence, word usage, sentence complexity, and grammar. Advanced models like Long Short-Term Memory (LSTM) networks have been effective in capturing the nuanced patterns in text data.

Project Objective

The primary objective of this project is to develop a state-of-the-art AES system using a hybrid model that incorporates neural networks. This system aims to improve upon existing methods by utilizing advanced feature extraction techniques and leveraging large-scale essay datasets.

Methodology

1. Data Collection and Preprocessing

• Datasets: Utilize publicly available datasets such as the Automated Student Assessment Prize (ASAP) dataset for training and evaluation. The dataset can be accessed at Kaggle.
• Feature Extraction: Extract relevant textual features such as syntactic structures, semantic coherence, and lexical diversity.

2. Model Architecture

• Neural Network Model: Implement a neural network model with LSTM layers to capture sequential dependencies in text.
• Attention Mechanism: Incorporate an attention mechanism to focus on critical parts of the essay that are most indicative of quality.

3. Training and Evaluation

• Training: Use a loss function suitable for regression tasks to train the model with backpropagation.
• Evaluation Metrics: Measure performance using metrics such as quadratic weighted kappa, accuracy, and mean squared error.

Expected Outcomes

The proposed system is expected to achieve higher accuracy in essay scoring compared to traditional methods. By utilizing deep learning techniques and attention mechanisms, the system should effectively handle variations in writing styles across different prompts and topics.

Conclusion

This project aims to advance the field of automated essay scoring by developing a cutting-edge system capable of accurately evaluating essays. The integration of neural networks and attention mechanisms is anticipated to provide significant improvements in performance.

For further details on related research, please refer to the paper "An Automated Essay Scoring Systems: A Systematic Literature Review," available at Springer.